|USFDC Home | USF Electronic Theses and Dissertations||| RSS|
This item is only available as the following downloads:
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam 22 Ka 4500
controlfield tag 007 cr-bnu---uuuuu
008 s2010 flu s 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0004582
Contribution to and use of online knowledge repositories :
b the role of governance mechanisms
h [electronic resource] /
by Varol Kayhan.
[Tampa, Fla] :
University of South Florida,
Title from PDF of title page.
Document formatted into pages; contains X pages.
Dissertation (PHD)--University of South Florida, 2010.
Includes bibliographical references.
Text (Electronic thesis) in PDF format.
Mode of access: World Wide Web.
System requirements: World Wide Web browser and PDF reader.
ABSTRACT: Drawing upon the concept of governance, this dissertation refers to the two most commonly employed mechanisms that ensure high quality knowledge in electronic repositories as expert-governance and community-governance. In three related but distinct essays, the dissertation examines the governance concept, and investigates contributing knowledge to and using knowledge from electronic repositories governed by these two mechanisms. The first essay sets the conceptual foundations of knowledge governance in repositories, and examines the salient aspects of expert- and community-governance that contribute to knowledge quality. The essay adopts an interpretive research methodology and analyzes empirical data collected from a range of organizations using interviews and online questionnaires. Findings suggest that executing governance functions thoroughly, experts' credibility, and experts' ownership of content contribute to knowledge quality in expert-governed repositories; and executing governance functions continuously and by a diverse set of members, and members' involvement in governance contribute to knowledge quality in community-governed repositories. The second essay investigates the factors that influence individuals to make voluntary contributions to expert- and community-governed repositories. This essay employs the same research methodology used in Essay I and suggests that personal benefits is a stronger motivator for contributing to expert-governed, and reciprocity is a stronger motivator for contributing to community-governed repositories when these two repositories are implemented on an individual basis in organizational settings. When the two repositories are implemented simultaneously, two sets of factors influence contribution behaviors: knowledge-based factors include the type, formality, and sensitivity of knowledge; and need-based factors include the need for collaboration, expert validation, and recognition. The third essay investigates knowledge use from expert- and community-governed repositories using a positivist perspective. It conducts a controlled experiment drawing upon elaboration likelihood model, and finds that the credibility of a governance mechanism positively affects subjects' perceptions of knowledge quality as well as their intentions to use knowledge, which in turn affect their actual knowledge use. This essay also conducts within-subject comparisons using repeated measures ANOVA to shed light on subjects' perceptions of expert- and community-governed knowledge assets.
Advisor: Anol Bhattacherjee, Ph.D.
x Information Sys and Decision Sci
t USF Electronic Theses and Dissertations.
Contribution to and Use of Online Knowledge Reposit ories: The Role of Governance Mechanisms by Varol O. Kayhan A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Information Systems/ Decision Science s College of Business University of South Florida Major Professor: Anol Bhattacherjee, Ph.D. Rosann Collins, Ph.D. Christopher Davis, Ph.D. Neset Hikmet, Ph.D. Balaji Padmanabhan, Ph.D. Date of Approval: July 6, 2010 Keywords: governance, expert-governance, communitygovernance, knowledge contribution, knowledge use Copyright 2010, Varol O. Kayhan
i TABLE OF CONTENTS LIST OF TABLES .................................... ................................................... ...................... ivLIST OF FIGURES ................................... ................................................... ..................... viABSTRACT .......................................... ................................................... ........................ viiiINTRODUCTION ...................................... ................................................... ......................1ESSAY I: GOVERNANCE OF KNOWLEDGE REPOSITORIES: A CONCEPTUAL FOUNDATION ............................. ...............................................5Introduction ...................................... ................................................... .....................5Overview of KM and Basic Concepts ................. ................................................... .6What is knowledge? ................................ ................................................... ..6Taxonomies of knowledge ........................... ..............................................10Knowledge Management .............................. .............................................11The Concept of Governance ......................... ................................................... ......13Governance of Knowledge in Repositories ........... ................................................17Governance in KM: Prior Research .................. ................................................... ..23Research Methods .................................. ................................................... .............26Grounded theory ................................... ................................................... ..28Open coding ....................................... ............................................29Axial coding ...................................... .............................................31Selective coding .................................. ...........................................33Data collection ................................... ................................................... .....34Sample characteristics ............................ ................................................... .35Participants in the first phase ................... ......................................35Participants in the second phase .................. ..................................37Data Analysis ..................................... ................................................... .....40Findings........................................... ................................................... ....................44Factors that contribute to knowledge quality ...... .......................................44Assessment of knowledge quality ................... ...........................................55Trustworthiness of findings ....................... ................................................... .........59Discussion ........................................ ................................................... ...................63Key findings ...................................... ................................................... ......63Limitations of the study .......................... ................................................... 66Theoretical implications........................... ..................................................6 8
ii Practical implications ............................ ................................................... ..73ESSAY II: USERS MOTIVATIONS TO CONTRIBUTE TO EXPER TAND COMMUNITY-GOVERNED REPOSITORIES ................... ...............................76Introduction ...................................... ................................................... ...................76Prior Research .................................... ................................................... .................77Research Methods .................................. ................................................... .............84Data collection procedure ......................... .................................................85Sample characteristics ............................ ................................................... .88Data analysis ..................................... ................................................... ......88Findings........................................... ................................................... ....................93Existence of one governance mechanism ............. .....................................93Organizational benefits ........................... .......................................95Reputation ........................................ ..............................................96Altruism .......................................... ...............................................98Organizational rewards ............................ ......................................99Personal benefits ................................. .........................................100Reciprocity ....................................... ............................................102Codification effort ............................... .........................................103Lack of expertise ................................. .........................................104Risk of duplication ............................... ........................................105Existence of two governance mechanisms ............ ..................................106Suggestions/ideas ................................. ........................................108Sensitivity of knowledge........................... ...................................111Formality of contributions ........................ ...................................113Need for collaboration ............................ .....................................116Need for expert validation ........................ ...................................118Need for recognition .............................. ......................................120Trustworthiness of Findings ....................... ................................................... ......123Discussion ........................................ ................................................... .................123Key findings ...................................... ................................................... ....123Limitations of the study .......................... .................................................12 6Theoretical implications........................... ................................................127Practical implications ............................ ................................................... 131ESSAY III: THE ROLE OF GOVERNANCE MECHANISMS IN USI NG KNOWLEDGE FROM REPOSITORIES ....................... ....................................134Introduction ...................................... ................................................... .................134Prior Research .................................... ................................................... ...............135Theory and Research Model ......................... ................................................... ....141Elaboration Likelihood Model ...................... ...........................................141Research Model .................................... ................................................... 143Research Methods .................................. ................................................... ...........147Subjects and Design ............................... ..................................................1 47Experimental setup................................. ..................................................1 50Procedure ......................................... ................................................... .....151Operationalization of Constructs .................. ...........................................154
iii Findings........................................... ................................................... ..................157Pilot experiment .................................. ................................................... ..157Experiment ........................................ ................................................... ....159Outlier analysis .................................. ..........................................160Manipulation check ................................ ......................................162Order effects...................................... ...........................................163Scale validity .................................... ............................................167Hypotheses testing ................................ .......................................172Post-hoc analysis ................................. ................................................... ..177Assumptions ....................................... ................................................... ...185Discussion ........................................ ................................................... .................189Key findings ...................................... ................................................... ....189Limitations of the study .......................... .................................................19 1Theoretical implications........................... ................................................193Practical implications ............................ ................................................... 198CONCLUSION ........................................ ................................................... .....................201REFERENCES ........................................ ................................................... .....................205APPENDICES ........................................ ................................................... ......................216Appendix A ........................................ ................................................... ...............217Appendix B ........................................ ................................................... ...............221Appendix C ........................................ ................................................... ...............222Appendix D ........................................ ................................................... ...............223Appendix E ........................................ ................................................... ...............225Appendix F......................................... ................................................... ...............226 ABOUT THE AUTHOR .................................. ................................................. E ND PAGE
iv LIST OF TABLES Table 1. Differences between data, information, and knowledge ........................................ 7Table 2. Taxonomies of Knowledge .................. ................................................... .............10Table 3. Knowledge Management Processes............ ................................................... ......14Table 4. Breakdown of participants in the first pha se ................................................ .......37Table 5. Participant comments for quality implicati ons of expert-governance .................41Table 6. Concepts and categories identified for exp ert-governance ..................................42Table 7. Concepts and categories identified for com munity-governance ..........................49Table 8. Concepts and factors identified for qualit y ................................................. .........56Table 9. Dimensions of knowledge quality identified in the literature ..............................58Table 10. A sample of independent variables investi gated in the literature ......................81Table 11. Processes identified in the literature .. ................................................... .............82Table 12. Dependent variables investigated in the l iterature ......................................... ....83Table 13. Types of repositories studied and their g overnance mechanisms ......................84Table 14. Participant comments for providing contri butions to expert-governance .........89Table 15. Concepts and categories identified for en ablers of expert-governance .............90Table 16. Concepts and categories identified for ch oice of governance mechanism ......108Table 17. Types of repositories studied and their g overnance mechanisms ....................140
v Table 18. Measurement Items ....................... ................................................... ................155Table 19. Pilot experiment descriptive statistics.. ................................................... .........158Table 20. Distribution of subjects within groups .. ................................................... ........161Table 21. Descriptive statistics of measurement ite ms ................................................ ....162Table 22. Results of the manipulation check ....... ................................................... .........163Table 23. Order effects ........................... ................................................... ......................166Table 24. Factor loadings of items used for the exp ert-governed page ...........................168Table 25. Factor loadings of items used for the com munity-governed page ...................169Table 26. Composite reliability, AVE, and correlati ons for the expert-governed page ............................................. ................................................... ..................170Table 27. Composite reliability, AVE, and correlati ons for the communitygoverned page .................................... ................................................... ...........171Table 28. Comparison of interaction models with mai n effects models ..........................173Table 29. P-values of Boxs homogeneity of covarian ces test ........................................18 7Table 30. P-values of Levenes homogeneity of varia nces test .......................................18 8
vi LIST OF FIGURES Figure 1. Different types of governance mechanisms ................................................... ....22Figure 2. Second-phase participants work experienc e ................................................. ....39Figure 3. Hierarchical structure of constructs for expert-governance ...............................43Figure 4. Screenshot of an example question ....... ................................................... ..........86Figure 5. Screenshot of an example interview questi on in the second phase ....................88Figure 6. Hierarchical structure of categories .... ................................................... .............92Figure 7. Comparison of factors identified for expe rtand community-governed repositories ...................................... ................................................... .................94Figure 8. Choice of governance mechanisms ......... ................................................... ......107Figure 9. Research Model for Essay III ............ ................................................... ............144Figure 10. Experimental design .................... ................................................... ................149Figure 11. Mean credibility scores of governance me chanisms in the pilot experiment ...................................... ................................................... .............159Figure 12. Effects of counterbalancing on measureme nt ................................................ .165Figure 13. Parameter estimates of expert-governance model ..........................................17 4Figure 14. Parameter estimates of community-governa nce model ..................................176Figure 15. Repeated measures ANOVA for credibility of governance mechanism ........179Figure 16. Repeated measures ANOVA for knowledge qu ality......................................182
vii Figure 17. Repeated measures ANOVA for intention .. ................................................... 183Figure 18. Repeated measures ANOVA for knowledge us e ...........................................184Figure A 1. High credibility expert-governed page .................................................. .....217Figure A 2. Low credibility expert-governed page ................................................... ....218Figure A 3. High credibility community-governed p age ..............................................2 19Figure A 4. Low credibility community-governed pa ge ...............................................2 20Figure B 1. Instructions given to subjects ...... ................................................... ............221Figure C 1. The link of the first treatment provi ded to subjects ................................... .222Figure D 1. Sample comprehension questions relate d to the governance mechanism ................................... ................................................... ............223Figure D 2. Sample comprehension questions relate d to the information on a Web page .................................... ................................................... .............224Figure E 1. Measurement of knowledge use from the two pages..................................225Figure F 1. Interaction effects model for the exp ert-governed page .............................226Figure F 2. Interaction effects model for the com munity-governed page .....................227
viii Contribution to and Use of Online Knowledge Reposit ories: The Role of Governance Mechanisms Varol O. Kayhan ABSTRACT Drawing upon the concept of governance, this disser tation refers to the two most commonly employed mechanisms that ensure high quali ty knowledge in electronic repositories as expert-governance and community-gov ernance. In three related but distinct essays, the dissertation examines the gove rnance concept, and investigates contributing knowledge to and using knowledge from electronic repositories governed by these two mechanisms. The first essay sets the con ceptual foundations of knowledge governance in repositories, and examines the salien t aspects of expertand communitygovernance that contribute to knowledge quality. T he essay adopts an interpretive research methodology and analyzes empirical data co llected from a range of organizations using interviews and online questionn aires. Findings suggest that executing governance functions thoroughly, experts credibility and experts ownership of content contribute to knowledge quality in expert-governed repositories; and executing governance functions continuously and by a diverse set of members, and members involvement in governance contribute to knowledge quality in c ommunity-governed repositories.
ix The second essay investigates the factors that infl uence individuals to make voluntary contributions to expertand community-go verned repositories. This essay employs the same research methodology used in Essay I and suggests that personal benefits is a stronger motivator for contributing to expert -governed, and reciprocity is a stronger motivator for contributing to community-go verned repositories when these two repositories are implemented on an individual basis in organizational settings. When the two repositories are implemented simultaneously, tw o sets of factors influence contribution behaviors: knowledge based factors include the type, formality, and sensitivity of knowledge; and need based factors include the need for collaboration, expert validation, and recognition. The third essay investigates knowledge use from exp ertand communitygoverned repositories using a positivist perspectiv e. It conducts a controlled experiment drawing upon elaboration likelihood model, and find s that the credibility of a governance mechanism positively affects subjects perceptions of knowle dge quality as well as their intentions to use knowledge, which in turn affect t heir actual knowledge use. This essay also conducts within-subject comparisons using repe ated measures ANOVA to shed light on subjects perceptions of expertand community-g overned knowledge assets.
1 INTRODUCTION The number of organizations that implement knowledg e management (KM) systems to increase efficiency and effectiveness, a nd gain competitive advantage is on the rise (Davenport et al., 2008). Electronic reposito ries are an essential component of these systems since they build organizational memory and store knowledge assets for future use by organizational members (Alavi and Leidner, 2 001; Holzner and Marx, 1979; Huber, 1991). It has been widely acknowledged that knowledge transfer depends partly on the availability of high-quality knowledge in th ese repositories (Hansen et al., 1999; Pentland, 1995; Schuler, 1994; Wiig, 1997). Anecdo tal evidence suggests that organizations use two different approaches to satis fy this need. The first uses experts or supervisors as referees to vet users contributions made to repositories; the second uses a community of users to review, rate, or edit existin g contributions in repositories. The first approach is the most commonly used mechan ism, as expert validation has been around for centuries and is the predominan t approach for moderating the development and communication of new knowledge (Kro nick, 1990). An example repository that employs this approach is WebMD (htt p://www.webmd.com), which provides answers to health related problems. The r epository publishes contributions provided by physicians only after they are reviewed by an expert physician in that domain. The second approach is a more recent devel opment, owing its existence to advancements in technology. This is because it wou ld have been very difficult, if not
2 impossible, to use this approach without the featur es afforded by current technologies, especially those that are commonly associated with Web 2.0. An example knowledge repository on the Web that employs this approach is Wikipedia (http://www.wikipedia.com), which houses user-gener ated content on a variety of topics ranging from science to entertainment. Drawing upon the sociology literature, this dissert ation refers to these two approaches as expert-governance and community-governance respectively. Expertgovernance is similar to the centralized and hierar chical form of societal governance as experts enforce policies and procedures on contribu tors to increase the quality of knowledge in repositories. On the other hand, comm unity-governance is similar to the decentralized and autonomous form of societal gover nance as a community of users increases the quality of knowledge in repositories through collective effort. Although the use of expertand community-governance is prevalen t in many organizations, our understanding of them, and their contribution to th e process of governance an emerging and important concept in contemporary business is rather limited. The goals of this dissertation are to set the conceptual foundations of this new concept, distinguish between different forms of governance, and extend o ur understanding of knowledge contribution and knowledge use in the existence of expertand community-governance. The dissertation is structured in three related by distinct essays. The first essay, titled Governance of Knowledge Repositories: A Conceptual Foundation , develops the concept of knowledge governance in electronic repos itories, reviews critical KM literature, and discusses how the governance concep t fits the existing KM literature. This essay also examines the ways with which expertand community-governance improve
3 knowledge quality in organizational repositories us ing an interpretive paradigm. It uses grounded theory to analyze empirical data collected from a range of organizations, and proposes a number of significant relationships for expertand community-governance and criteria used to assess knowledge quality. The second essay, titled Users Motivations to Contribute to Expertand Community-Governed Repositories , adopts the same research methodology employed in the first essay, and aims to identify the factors t hat influence individuals to voluntarily make contributions to expertand community-governe d repositories used in organizations. This essay develops theoretical mod els and propositions for two different contexts, one in which organizations use only one t ype of repository (either expertor community-governed), and another in which both type s of repositories are used simultaneously. The third essay, titled The Role of Governance Mechanisms in Using Knowledg e from Repositories , examines the use of knowledge from expertand c ommunitygoverned repositories from a positivist perspective Drawing upon the elaboration likelihood model (Petty and Cacioppo, 1986a), this essay hypothesizes that the credibility of a governance mechanism influences individuals q uality perceptions and their intentions to use knowledge, which, in turn, affect s their actual knowledge use. To test these hypotheses, the essay reports a controlled ex periment where subjects are exposed to knowledge assets that are governed by either expert or community-governance with varying levels of credibility. The analysis is dee pened through repeated measures ANOVA to shed light on what transpires if individua ls are exposed to different forms of governance in a sequential manner.
4 The final section of the dissertation synthesizes t he contributions from the three essays. Following a brief a summary of the finding s, important implications of this dissertation for theory and practice are highlighte d.
5 ESSAY I: GOVERNANCE OF KNOWLEDGE REPOSITORIES: A CONCEPTUAL FOUNDATION Introduction The goals of this essay are to set the foundations of the governance concept, distinguish between different types of governance m echanisms used for organizational knowledge repositories, and focus on two commonly u sed mechanisms, expertand community-governance, to understand how (if ever) t hese mechanisms increase knowledge quality in electronic repositories. Ther efore, in addition to developing a conceptual foundation, this essay addresses the fol lowing research question: do expertand community-governance improve the quality of kno wledge in organizational knowledge repositories; and why, or why not? This essay is motivated by the fact that governance mechanisms, such as expertand community-governance, are used commonly in many organizations; however, neither practitioners nor academics are fully aware of thei r differences or their salient aspects that contribute to knowledge quality. For instance the traces of expertand communitygovernance can be observed in prior research (e.g., Alavi et al., 2006), popular press (e.g., Nevo et al., 2009), or industry reports (e.g ., McKinsey, 2008), although no one to the best of our knowledge has distinguished betwe en them or provided suggestions about how they improve knowledge quality. The exta nt literature lacks conceptual development in defining governance mechanisms. Con sequently, there are no well
6 developed explanations of how these mechanisms incr ease knowledge quality in repositories. This essay aims to address these gap s in the literature from an interpretive perspective. It uses a grounded theory approach to analyze the empirical data collected from professionals in a range of organizational set tings. The essay first defines and differentiates between different governance mechani sms, then identifies the salient aspects of the two mechanisms, expertand communit y-governance, that contribute to knowledge quality. The remainder of the essay proceeds as follows. In the next section, critical KM literature is reviewed. The following section surv eys the governance literature in sociology and extends the mechanisms used for socie tal governance to the context of KM. In the next section, prior research in knowled ge governance is examined providing a basis for distinguishing the concept of governanc e developed in this dissertation from earlier work in knowledge governance. The followin g section examines the research question posed in this essay, and presents the find ings about the aspects of expertand community-governance that contribute to knowledge q uality. The final section discusses the theoretical, practical, and research implicatio ns of this research. Overview of KM and Basic Concepts What is knowledge? The meaning of knowledge has led to many philosophi cal debates throughout the history beginning from the Greek era. The epistemo logical differences between philosophers have made it difficult to define knowl edge and therefore led researchers to define knowledge by distinguishing it from data and information (Alavi and Leidner, 2001; Nonaka, 1994; Polanyi, 1958).
7 It has been widely accepted that data comprises raw facts, unprocessed numbers, or observations about the states of the world; info rmation is processed data, or data that is given a purpose; and knowledge is authenticated inf ormation, or information that is given a context, interpretation, and meaning (Alavi and Leidner, 2001; Davenport, 1997; Dretske, 1981; Drucker, 1988; Machlup, 1980; Vance, 1997). This distinction creates a hierarchy, in which information is derived from dat a, and knowledge is derived from information. The differences between data, informa tion, and knowledge as suggested in the prior literature are summarized in Table 1. Data Information Knowledge Raw facts Unprocessed numbers Observations about the states of the world Processed data Data that is given purpose Authenticated information Information that is given meaning, interpretation, and context Table 1. Differences between data, information, and knowledge The following example illustrates the data-informat ion-knowledge hierarchy, and shows how they differ. There are many important fa ctors that determine the intensity of a hurricane, one of which is water temperature. Rese archers investigating the intensity of hurricanes in the Gulf of Mexico create data by measuring the water temperature in the gulf from many sensors at a given point in time. T he measurements (i.e., data) correspond to raw facts about or different states o f gulf water. If the researchers choose to categorize these measurements as to whether or n ot they are in the Loop Current (the circular stream of warm water in the Gulf of Mexico ), they creates information This is because the researchers process the data, and give it a purpose to communicate a certain message. If the researchers develop an understandi ng of how the temperature difference
8 between the Loop Current and the rest of the gulf w ater intensifies a hurricane, they create knowledge In this case, the researchers interpret the info rmation, and give it meaning and context. If a tropical storm is headed to the researchers town under unfavorable Loop Current temperatures, they will li kely start packing or seek shelter as they know that the storm will intensify to a (potentially po werful) hurricane. On the other hand, other people in the same town, looking at the same information may not take any action as they do not know the relationship between the Loop Current temperat ure and hurricane intensity. As seen in this example, information is derived fro m data, and knowledge is derived from information, creating a hierarchy. Ho wever, this example also supports the notion of reverse hierarchy advocated by Tuomi (199 9). Tuomi (1999) argues that knowledge must exist before individuals formulate i nformation, and formulation of information must exit before individuals collect a specific set of data. For instance, in the preceding example, if the researchers had no idea a bout the relationship between the Loop Current temperature and hurricane intensity, t hey neither would have categorized the data with respect to the Loop Current, nor woul d have measured the water temperature in the Gulf of Mexico. Tuomis (1999) reverse hierarchy has important impl ications for the field of information systems (IS). One of these is that kno wledge precedes information, and therefore, articulation of knowledge can result in creating information. For this reason, Tuomi (1999) argues that knowledge management syste ms can easily turn into information management systems if individuals fail to codify t he interpretation, meaning, or context of information. A solution to this prob lem is to have certain mechanisms in
9 place (such as governance mechanisms as described i n this essay) to ensure that interpretation, meaning, and context are codified i n repositories. The data-information-knowledge hierarchy is not the only way to define knowledge. Others perspectives exist in the litera ture, defining knowledge variously as a state of mind (i.e., experienced-based understandin g), an object (i.e., a thing that can be stored and manipulated), a process (i.e., practicin g an expertise), a condition for accessibility, or a capability (i.e., ability to ta ke future actions). Alavi and Leidner (2001) provide an insightful comparison of these conceptua lizations. It is important to note that the aim of this disser tation is not to reconcile the philosophical differences in the literature. Rathe r, the dissertation treats knowledge and information as similar, and differentiates both fro m data. While discussion in this dissertation concerns only knowledge and informatio n, the term knowledge is used hereafter to refer to both due to their interdepend ence. The challenge of this distinction has been raised before by Davenport (1997), who sta tes that the distinction is rather imprecise. The use of knowledge repositories in practice also makes the distinction irrelevant, as most repositories store not only ins ights gained from experience (which can be considered knowledge), but also contextualized a nd processed facts (which can be considered information). For example, it is very c ommon for consulting firms to use knowledge repositories to store best practices or l essons learned about a consulting job (i.e., knowledge) as well as tax rates or regulatio ns (i.e., information). For consistency, the term is knowledge is used throughout the disser tation to refer to both knowledge and information.
10 Taxonomies of knowledge The KM literature suggests that there are different types of knowledge. Some of the most commonly accepted taxonomies are presented in Table 2. The most popular of these is Nonakas (1994) tacit-explicit taxonomy. Drawing upon the work of Polanyi (1958), Nonaka (1994) states that explicit knowledg e is knowledge that is transmittable in formal, systematic language, while tacit knowle dge has a personal quality, which makes it hard to formalize and communicate (p.16). By nature, explicit knowledge can be codified, whereas tacit knowledge is difficult t o codify as it is rooted in experience, action, and involvement in a particular context. Types of knowledge Study Tacit Explicit Nonaka (1994); Polanyi (1958) General Context specific Zack (1999); Choudhury and Sabherwal (2001) Declarative Procedural Causal Analytic Zack (1999); Moorman and Miner (1998); Gottschalk (2000) Table 2. Taxonomies of Knowledge Besides the tacit-explicit taxonomy, researchers st ate that knowledge can be classified according to its specificity (Choudhury and Sabherwal, 2001; Zack, 1999); or the message it conveys (Gottschalk, 2000; Moorman a nd Miner, 1998; Zack, 1999). For example, knowledge can be general or context specif ic; or it may convey a declarative (i.e., describing something), a procedural (i.e., h ow something occurs or is performed), a causal (i.e., why something occurs), or an analytic message (i.e., outcome of applying declarative and procedural knowledge).
11 Extending the tacit-explicit taxonomy, Zander and K ogut (1995) state that tacitness (or codifiability) is only one of the dim ensions of knowledge, and knowledge has four other dimensions, namely teachability (i.e ., extent to which it can be taught), complexity (i.e., extent to which it draws upon dif ferent competencies), dependence (i.e., extent to which its creation depends on other peopl e or groups), and imitability (i.e., extent to which it can be copied). Knowledge Management Knowledge management (KM) is broadly defined as any capability or process that involves creating, capturing, storing, sharing and using knowledge in organizational settings (McAdam and McCreedy, 1999; Quintas et al. 1997; Swan et al., 1999; Wiig, 1997). It is noteworthy that this definition does not mention any information technology (IT), since IT plays a facilitating role in KM by enabling organizations to perform such processes (McAdam and McCreedy, 1999). Whether or not organizations use IT, the main purpose of any KM initiative is to leverage th e value of knowledge, thereby, improving organizational performance, maintaining s ustainability, and remaining competitive in market (Alavi and Leidner, 2001; Qui ntas et al., 1997; Swan and Newell, 2000). Prior literature does not consistently identify a s pecific set of processes that define or comprise KM. For example, Holzner and Marx (197 9) suggest that KM consists of five processes, namely construction, organization, storage, distribution, and application of knowledge. On the other hand, Huber (1991) argues that there are four processes that comprise KM: knowledge acquisition, information dis tribution, information interpretation, and organizational memory. Wiig (1 995) adopts another perspective,
12 suggesting that KM consists of four functional area s: governance functions, staff functions, operational functions, and realization o f value of knowledge. Alavi and Leidner (2001) offer some synthesis by combining th ese perspectives and propose that KM consists of four fundamental processes: (1) know ledge creation, which involves creating new knowledge or replacing existing knowle dge using organizations tacit and explicit knowledge; (2) knowledge storage and retri eval, which concerns storing organizational knowledge to, and retrieving it from organizations semantic and episodic memory; (3) knowledge transfer, which involves tran sferring individual explicit/implicit knowledge to group semantic/episodic memory; and (4 ) knowledge application, which involves applying knowledge to perform organization al tasks. These perspectives of KM are summarized in Table 3. This essay adopts Alavi and Leidners (2001) perspe ctive, and suggests that KM consists of knowledge creation, knowledge storage, knowledge transfer, and knowledge application processes. An important question that arises from this perspective is: where does the governance concept, and particularly knowledge governance in electronic repositories fit in KM? This question can be addressed in two different ways: (1) governance can be treated as a sub-process and incl uded under each major process (for example, the four processes can each have sub-proce sses called governance of knowledge creation, governance of knowledge storage, governan ce of knowledge transfer, and governance of knowledge application); or (2) govern ance can be treated as a standalone (i.e., fifth) process incorporating any governancerelated sub-processes. This essay adopts the latter approach, since recent research h as identified an overarching process KM governance (Foss, 2007; Schroeder and Pauleen, 2007). This e ssay considers
13 knowledge governance in electronic repositories as one of the sub-processes of KM governance. Another important question is: which processes do g overnance mechanisms impact the most? Since governance mechanisms striv e to increase the quality of knowledge assets, it is expected that they are most salient during knowledge codification, and knowledge retrieval. This suggests that govern ance of repositories is important at the input and output stages of knowledge management. Input corresponds to knowledge contribution, where individuals codify their tacit knowledge int o explicit for storing in organizational repositories. On the other hand, ou tput corresponds to knowledge use where individuals retrieve explicit knowledge from organizational repositories to be used in performing organizational tasks (Nonaka, 1994). The Concept of Governance Kooiman and Bavinck (2005) define governance as th e whole of public as well as private interactions taken to solve societal pro blems and create societal opportunities (p.17). According to this conceptualization, gover nance can be considered arrangements (or mechanisms ) that can solve problems faced by a group of indiv iduals, collective, community, or society (Kooiman, 1999). The sociolo gy literature provides a comprehensive exposition of such mechanisms, two of which are hierarchical control and community-governance.
14 Study Knowledge Management Processes Holzner and Marx (1979) Construction: Developing and adding new knowledge to the existing stock of knowledge Organization: Classifying and integrating existing knowledge, or relating it to one another Storage: Storing knowledge to develop organizational memory Distribution: Distributing knowledge to places where it is needed Application: Applying knowledge to perform organizational tasks Huber (1991) Acquisition: Obtaining knowledge (either from acquiring or creating it) Distribution: Shared information by others Interpretation: Giving a distributed information a common interpretation Memory: Storing knowledge for future use Wiig (1995) Governance functions: Monitoring and facilitating knowledge related processes Staff functions: Establishing and updating knowledge infrastructure Operational functions: Creating, renewing, building, and organizing knowledge assets Realization of value of knowledge: Distributing and applying knowledge Alavi and Leidner (2001) Creation: Creating new knowledge using organizations tacit/explicit knowledge Storage/retrieval: Storing knowledge to develop semantic/episodic organizational memory, and retrieving knowledge from these memories Transfer: Transfer of individual explicit/implicit knowledge to group semantic/episodic memory Application: Applying knowledge to perform organizational tasks Table 3. Knowledge Management Processes
15 Hierarchical control represents the classical top-d own approach between governors (i.e., state) and the governed (i.e., cit izens), in which the state imposes rules and policies on citizens to provide services. It i s in the best interest of citizens to abide by the rules, because failure to do so can result i n punishment. The states coercion through policies is legitimate, and performed by ci vil servants. The fundamental motivations of civil servants to enforce these poli cies are career advancement and the bureaucratic stability provided by the state. Hier archical control can achieve its intended goals if the state can provide its citizens with se curity, equal and predictable treatment, and efficient mobilization of resources (Streeck an d Schmitter, 1985). However, hierarchical control can also suffer from certain l imitations such as creating tensions between the state and citizens over the privileges of incumbents or the obligations imposed on citizens (Streeck and Schmitter, 1985). Further, hierarchical control is considered to be more susceptible to moral hazard a nd adverse selection problems as it is difficult for civil servants to monitor all citizen s (Bowles and Gintis, 2002). A second mode of governance is community-governance where citizens take care of themselves and solve problems on their own rathe r than relying on the state. Community-governance occurs through individuals au tonomous and voluntary efforts to deal with societal problems. As community-governan ce takes advantage of the information dispersed among citizens, it is less su sceptible to the problems of moral hazard and adverse selection that plague hierarchic al control (Bowles and Gintis, 2002). Community-governance is usually preferred over hier archical control if the context is diverse, complex, and dynamic (Kooiman, 1999). Thi s is because, in such a context, there is no single person, group, or organization t hat has the power, authority, knowledge,
16 or resources to solve problems (Bryson and Crosby, 1993). Kooiman (1999) proposes that community-governance requires three essential components: images instruments and actions Images represent the guiding light of governan ce (e.g., a shared goal), and concern individuals visions, knowledge, facts, jud gments, ends, goals, etc. Instruments are tools that enable individuals to enact their im ages. They can be either soft (such as information, peer pressure, bribe, etc.), or hard (such as covenants, agreements, etc.). Actions are putting instruments into effect, and th ereby implementing images. Community-governance has its own share of problems compared to hierarchical control. For instance, it may lead to the formatio n of cliques, which can alienate community members especially if a core group of mem bers treat others as foreigners (Streeck and Schmitter, 1985). This, in turn, can cause the alienated members to leave the community, which makes the community more homog eneous, stripping it of the benefits of diversity, and even causing groupthink (Bowles and Gintis, 2002; Janis, 1982). Hierarchical control and community-governance are n ot the only mechanisms employed in societies, as markets or associations can also be used to tackle societal problems (Streeck and Schmitter, 1985). In markets political parties represent electoral voice and compete with one another to provide servi ces to, and solve problems of citizens. Parties develop and pitch policies tha t outline which problems will be solved and how, and then try to maximize their electoral v ote to put their policies in place. In contrast, associations involve actors, such as orga nizations, that solve their problems through concertations or negotiations that are impl emented as pacts. These pacts allow actors to recognize each others status and entitle ments in pursuing their individual
17 interests, and use collective effort to reach commo n goals. This essay (and dissertation) focuses on hierarchical control and community-gover nance, since they are the two most relevant mechanisms to the concept of repository go vernance in the context of KM. The concepts of hierarchical control and communitygovernance has already been extended to the organizational context to explain t he development of workflow formalization (Adler and Borys, 1996). Adler and B orys (1996) argue that the problem of formalizing process workflows (i.e., developing rules, procedures, and instructions for workflows) can be addressed using two approaches: coercive and enabling bureaucracy. Coercive bureaucracy corresponds to hierarchical control, where supervi sors design procedures and enforce. Subordinates are required to implement these procedures without any deviations, and are not expected to ada pt them. Rules and procedures are rigid since the fundamental assumption is that supe rvisors prescribe, subordinates implement, and supervisors authorize deviations if needed. On the other hand, enabling bureaucracy corresponds to community-governance, where procedures are not designed exclusively by su pervisors, but also with the autonomous and voluntary participation of subordina tes. Subordinates are still required to implement procedures, but they also deal with co ntingencies and seek avenues for adaptation. Rules and procedures are flexible and can be overridden if deemed necessary. Governance of Knowledge in Repositories The concept of societal governance is relevant to K M, because governance, by definition, helps solve problems that are of inte rest to societies, organizations, or a group of individuals. Since increasing the quality of knowledge in electronic repositories
18 is a salient issue for many organizations, the conc ept of governance promises to be useful for KM. Before elaborating further on idea of knowledge gov ernance in electronic repositories, it is important to define this new te rm and identify different forms of governance in KM. By drawing upon the definition o f information technology (IT) governance proposed by the Information Technology G overnance Institute (ITGI, 2003, http://www.itgi.org), this dissertation defines the governance of knowledge in electronic repositories as the set of responsibilities and practices designed to increase the quality of knowledge in electronic knowledge repositories These responsibilities and practices can be exercised using different forms of governance (h ereafter referred to as governance mechanisms ). Organizations can employ many different governa nce mechanisms in an effort to increase knowledge quality in their repos itories. To identify some of these mechanisms, we turn to the definition of governance is sociology. Governance is defined as the whole of public as we ll as private interactions taken to solve societal problems and create societa l opportunities (Kooiman and Bavinck, 2005, p.17, emphasis added). This definit ion suggests that an important aspect of governance is interactions because in order to achieve a desired outcome or solve a societal problem, governors and the governed need t o interact with each other. Through interaction, governors communicate the rules and po licies to the governed, and the governed provide feedback to the governors about th eir implications. The feedback provided by the governed helps the governor make mo difications to the rules and policies if necessary. The sociology literature suggests th at governance mechanisms that lack adequate interactions between governor and governed are less likely to achieve their
19 intended goals, because interactions reinforce the influence of the governor on the governed (Kooiman, 1999). For example, a driver pu lled over by a police officer, or cited for careless driving will be more likely to f ollow traffic rules even if there is no possibility of being pulled over or cited again. F or this reason, governance is not merely something governors do, but a quality of the totali ty of the interactions between those governing and those governed (Kooiman and Bavink 2 005, p,19). Similarly, interactions in KM play an important rol e for instantiating different types of governance mechanisms. There are two diff erent types of interactions in KM: (1) interactions between the governor and the gover ned (governor-governed interactions); and (2) interactions between the governor and the c ontent (governor-content interactions). In governor-governed interactions, the governor pro vides feedback to the governed to help them make high quality contributio ns to the repository. For example, a designated group of experts review knowledge contri butions and provide feedback to contributors to help increase content quality. Thi s type of interaction occurs before the submission is published in the repository. This di ssertation refers to the governance mechanism that uses this type of interaction as expert governance. Expert-governance corresponds to the hierarchical mode of governance described in the sociology literature, where experts or supervisors act as referees, and a ccept or reject contributions made to a knowledge repository. If submissions are below par experts may require authors to revise their submissions before publishing them in the repository. Any revisions to published content can also be subjected to a simila r process, where experts or supervisors evaluate change requests and allow changes that are deemed necessary. From a
20 technological design perspective, expert-governance uses technology to disseminate high quality content. After a submission is published, technology does not allow users to interact with one another or to provide feedback to the original contributor. For this reason, expert-governance provides unidirectional i nformation flow between users and repositories. The second type of interaction that is prevalent in KM is governor-content interaction, where governors interact with the publ ished content in electronic repositories rather than the contributors. In this case, contri butors to a repository act as governors and edit the existing content, or provide comments or r atings to either increase or assess the content quality. This type of interaction is diffe rent from governor-governed interaction, because unlike experts, contributors do not enforce the author to make changes to the content, but rather change the content themselves ( or provide comments or ratings). This dissertation refers to the governance mechanism tha t uses this type of interaction as community governance where community refers to a group of individuals who share the same responsibilities, who work in the same domain, or who are contributors to the same business process in the same organization. The tec hnological design of communitygovernance is fundamentally different from expert-g overnance in that communitygoverned repositories must provide technological fe atures that allow contributors to the repository to interact with the content through rev iewing, editing, rating, etc. Therefore, technology not only helps disseminate high quality content, but also enables members to interact with the published content through differe nt types of design features. It is important to note that there can be other typ es of governance mechanisms that rely on governor-content interaction. For exa mple, organizations can implement
21 agent-based systems, where software agents interact with content published in repositories by collecting meta-data through crawli ng. In this case, agents do not necessarily increase the content quality, but help organizations improve the overall quality of a repository (by indexing, classifying, or tagging knowledge assets), which help knowledge users retrieve the most relevant (an d therefore, highest quality) content from the repository. This dissertation refers to t his type of governance mechanism as auto-governance. An example of auto-governance is the Google search engine, which uses Web crawlers to collect data about Web pages, applies indexing and classification techniques to the crawled data, then uses a proprie tary page rank algorithm to identify the most relevant information on the Web. There is a third mechanism, besides communityand auto-governance, that relies on governor-content type of interaction. In this m echanism, governors interact with only their own contributions rather than others. This could arise from either certain restrictions imposed on contributors to the reposit ory (such as not being allowed to edit or provide comments or ratings on others contribution s), or from social norms in the organization. This type of mechanism is referred t o as self governance in this dissertation. In repositories that employ this mec hanism, content is usually accessible by everyone, but only corresponding contributors are r esponsible for increasing the quality of their contributions. For example, a file sharin g server, or static intranet pages for knowledge sharing can be considered self-governed r epositories as only the original contributors may have the permission to update thei r contributions. In summary, it is possible to identify four differe nt governance mechanisms that are instantiated through two types of interactions: governor-governed and governor-
22 content. These four types of mechanisms are presen ted in Figure 1 below. Of these four mechanisms, this dissertation focuses specifically on expertand community-governance, since they are the two most commonly used mechanism s in organizations. Governor-governed Interaction Governor-content Interaction Expertgovernance Communitygovernance Autogovernance Selfgovernance Figure 1. Different types of governance mechanisms Having classified the governance mechanisms used in electronic repositories, there are two issues that need further clarificatio n. First, the four types of governance mechanisms identified are not mutually exclusive: t here can be hybrid mechanisms. For instance, organizations can use both expertand co mmunity-governance by having a designated group of experts review initial submissi ons made to a repository, then allowing contributors to the repository provide rat ings or comments about these submissions once they are published. Investigation of such hybrid mechanism is beyond the scope of this dissertation, since the goal of t his research is to examine the specific differences between expertand community-governanc e. Second, this dissertation rests on the assumption that governance mechanisms are us ed only to increase the quality of knowledge in electronic repositories as opposed to promoting any political agenda. Since governance mechanisms, especially expertand commu nity-governance, are a manifestation of organizational power, it is possib le to use governance mechanisms to exert influence on organizational members. For exa mple, expert-governance can be used to censor certain types of knowledge (such as organ izational, departmental, or managerial
23 failures or weaknesses) from organizational members Such censorship might be prompted by concerns that these types of knowledge might jeopardize authority or legitimacy in an organization. Censorship might oc cur unconsciously, through tacit screening by experts, or explicitly (and consciou sly) by or under the direction of senior managers. Consequently, regardless of who c ensors, contributors to the repository might be intentionally exposed to only certain type s of knowledge. Similarly, community-governance can be used as a tool to play politics, or change the power dynamics in an organization. For example, individu als might undermine the validity and quality of certain types of knowledge (such as thos e that advocate an innovation or process design) especially if a conflict of interes t exists. It is important to note that this dissertation espouses a rational perspective that governance mechanisms are used to increase quality of knowledge in electronic reposit ories, rather than promoting any political agenda. This is a necessary limitation o f the epistemological position adopted in order to maintain focus on the research question an d the validity of the empirical analysis it prompts. Governance in KM: Prior Research Governance of knowledge in electronic repositories as discussed above has not been conceptualized in the KM literature, despite t he fact that KM research has frequent references to knowledge governance. Various resear chers have alluded to KM governance in recent years variously as a set of activities, p olicies, or procedures that control, coordinate, and facilitate the knowledge m anagement processes in organizations (Foss, 2007; Schroeder and Pauleen, 2007). This la ck of cohesion presents an opportunity to categorize studies into different gr oups according to their specific focus.
24 One group of studies investigates the governance of knowledge transfer, and sheds light on how knowledge transfer is controlled and facilit ated within and between organizations. For example, job design, reward sys tems, information systems, online communities, property rights, and patents are consi dered different forms of governance mechanisms that facilitate knowledge transfer betwe en and within firms (Foss, 2007; Grandori, 2001; Krafft and Ravix, 2008). Among the studies that focus on the governance of knowledge transfer within firms, Davenport and colleagues (Davenport, 1997; Davenport et al., 1992; Strong et al., 2008) examine different mechanisms that regulate inter-departmental flow of knowledge. The y suggest that organizations adopt various mechanisms depending on the degree to which employees perceive information as a source of power. Accordingly, five types of g overnance mechanisms, namely technocratic utopianism, monarchy, federalism, feud alism, and anarchy explain how knowledge transfer takes place. While technocratic utopianism represents the ideal that knowledge flows freely in organizations (if there e xists a carefully planned IT infrastructure), the other four types of mechanism (from monarchy to anarchy) are conceptualized as a continuum of local versus centr alized control of knowledge transfer. For instance, in monarchy, a powerful executive (su ch as the CEO) dictates the rules for transfer of knowledge, whereas in anarchy there are no formal rules as individuals advocate for their own needs. In his later work, D avenport (1997) adds to this typology a market-based mechanism, where knowledge transfer is controlled through market prices. Among the studies that focus on the governance of k nowledge transfer between organizations Mu et al. (2008) considers social capital a governance mechanism, and argues that weak ties help develop initial relation ships between organizations, and trust-
25 based strong ties accelerate high-quality and finegrained knowledge transfer. Similarly, Choi et al. (2005) argue that three mechanisms, nam ely market-based governance, entitlement governance, and gift governance, are sa lient to knowledge transfer between organizations. In market-based governance, knowled ge transfer takes place at market prices; in entitlement governance, organizations en force their right to obtain knowledge from other organizations; and in gift governance, k nowledge transfer takes place based on the goodwill and trust of interacting organizati ons. The governance of knowledge transfer is not the onl y focus in the literature. Researchers also focus on the governance of KM effo rts by developing and implementing new KM strategies (e.g., Zyngier et al., 2006); and by defining the roles of KM leaders (e.g., Chourides et al., 2003) or community sponsor s or facilitators (Lank et al., 2008). Although the above studies provide useful insights about how organizations can go about managing knowledge transfer between and wi thin firms, they do not clearly articulate the concept of governance. They inform us of different mechanisms that control, coordinate, and facilitate knowledge trans fer, and make policy-based suggestions about various KM strategies as well as roles of KM stakeholders. However, the extant literature falls short of clearly defining the conc ept of knowledge governance we propose, which addresses the quality of knowledge stored in electronic repositories. One exception is Neus and colleagues (Neus, 2001; Neus and Scherf, 2004), who discuss traditional and collaboration-oriented mechanis ms as alternative ways to manage knowledge in repositories. However, rather than ma king a distinction between the two or explaining the ways with which each mechanism impro ves knowledge quality, they make a rather deterministic assessment and suggest that collaboration-oriented techniques (such
26 as wikis) are superior to traditional systems in c reating, sharing, and managing information. Further, they rely on observational a nd anecdotal evidence with very little clarity about the concept of governance. There is a clear need for conceptual development in this area that will not only extend the boundaries of the current discourse on governance, but will also pave the way for the d evelopment of new theories and frameworks that will enrich insights into knowledge governance in repositories. Among the governance mechanisms described earlier, expertand communitygovernance are being used widely in many organizati ons. However, prior research neither examines whether or not these mechanisms im prove knowledge quality, nor does it provide much insight into the aspects that contr ibute to quality. Therefore, as the first step of the investigation into governance mechanism s, this essay explores the effects of expertand community-governance on knowledge quali ty, and identifies salient aspects that improve quality. The next section discusses t he research methods used to achieve the goals of this essay. Research Methods Before describing the research methods employed in this essay, it is imperative to clarify some of the research methods terminology an d understand the differences between terms such as quantitative and qualitative research and the positivist and interpretive paradigms. Qualitative research involves the use of qualitative data, such as interviews, documents, and participant observation data, to und erstand and explain social phenomena (Myers, 1997, p.241). Qualitative resea rch is different from quantitative research in that quantitative research tries to qua ntify textual data into numbers (using, for example, Likert scales), whereas qualitative re search uses textual data as-is (in the
27 form of utterances or sentences) to capture the soc ial and institutional context of a natural setting (Kaplan and Maxwell, 1994). While the terms qualitative and quantitative resear ch relate to the type of data, the terms positivist and interpretive concern the epist emological assumptions being made for conducting social-science research. The positivist paradigm assumes that there is an objective reality out there, and it can be investig ated by testing hypotheses derived from a priori theories. On the other hand, the interpretive par adigm assumes that there is no objective reality, but the reality can be accessed or is constructed using language, consciousness, and shared meaning in a given contex t. Instead of testing hypotheses derived from prior theories, interpretive research tries to construct a different understanding and reality for each social and insti tutional context. The research methods used and the epistemological p aradigm adopted are not codependent. For example, it is possible to conduct qualitative research using either positivist or interpretive paradigms (Myers, 1997). Further, different types of research methodologies can be used for each approach accordi ng to the degree to which they serve the purposes of that approach. Methodologies inclu de grounded theory, ethnography, ethnomethodology, action research, and case study ( Myers, 1997; Strauss and Corbin, 1998). It is important to note that the type of me thodology is also partly independent of the type of paradigm and the type of research being conducted. For example, case study or action research can be used to conduct qualitati ve research using either a positivist or an interpretive paradigm. Having clarified some of the ambiguities surroundin g research terminology, it should be noted that this essay adopts an interpret ive perspective to conduct qualitative
28 research using grounded theory as a basis to the re search questions. The choice of paradigm was motivated by the dearth of a priori theories in the literature suited to the research question. Further, the qualitative nature of the research helps capture the social context in which governance mechanisms are investig ated, which is the central focus of the research question. The choice of grounded theo ry as the research methodology was also motivated by alignment with the question focus : (1) grounded theory emphasizes the importance of researchers immersion in data as muc h as other methods, (2) grounded theory allows the use of existing theoretical knowl edge, as opposed to suspending or ignoring it, to develop and enrich new theories (Gl aser, 1978), and (3) grounded theory leverages the strengths of both positivistic and in terpretive approaches in building new theories (Charmaz, 2000). Grounded theory involves the use of different types of tools and techniques for analyzing data and constructing new theories. The next subsection provides a brief description of grounded theory and its tools for data analysis. The following subsection explains the data collection t echniques used for this study. Following a description of the sample characteristi cs in the next subsection, the final subsection demonstrates how the data collected from participants were analyzed. Grounded theory Grounded theory is an inductive, theory discovery methodology that allows the researcher to develop a theoretical account of the general features of a topic while simultaneously grounding the account in empirical o bservations or data (Glaser and Strauss 1967). (Martin and Turner, 1986, p.141). Grounded theory is considered a research method as opposed to a coding procedure (M yers, 1997; Strauss and Corbin, 1998), because it induces researchers to ground new theories in empirical data through a
29 systematic analysis besides mere coding. Compared to hypothesis testing that deduces new theories from existing ones using the positivis t paradigm, grounded theory allows theories to emerge from the data through systematic analysis. This ensures that researchers construct the reality in a given contex t rather than allowing the existing theories to impose a certain external reality in th at context. The core of grounded theory lies in the use of thre e coding techniques, namely open, axial, and selective coding, that provide res earchers with the analytical tools for handling, examining, and making sense of raw data c ollected from participants. These techniques lead to theory building by allowing rese archers to identify concepts that are salient to the participants and thus the building b locks of theories. Below, the open, axial, and selective coding techniques are discusse d in depth. Open coding In general terms, open coding concerns opening up the data and exposing what is hidden inside. The main focus is to identify, u ncover, and name new concepts. Strauss and Corbin (1998) define a concept as a labeled ph enomenon (p.103) that represents an event, object, action or interaction. Once concept s are identified, they become meaningful entities for researchers to focus their attention on, and ask questions about. Questions about and answers elaborating these conce pts help researchers establish relationships that ultimately evolve into propositi ons or hypotheses, explaining why certain things happen the way they were observed in a given context. In order to identify concepts, open coding starts w ith breaking the data into small parts, and then examining each part to identify dis crete events, incidents, ideas, actions,
30 and interactions. After they are identified, conce pts can be named using two different approaches: (1) using the imagery or the meaning ea ch concept evokes in the researcher, (2) using the participants own naming convention ( which is referred to as in vivo codes ; [Glaser and Strauss 1967]). Following the identification of concepts, it is imp erative to identify the recognizable properties (or characteristics) of eac h concept such as its size, color, or capability. This is essential in order to further group similar (or relevant) concepts into more abstract categories. Categories are the building blocks of theories, a nd represent constructs Developing categories is important, because they reduce the amount of concepts the researcher needs to work with during d ata analysis. Categories should be named carefully: names should evoke imagery or mean ing quickly for the participant. It is also appropriate to use names from the existing literature particularly when researchers aim to extend current theories. However, caution n eeds to be used with using existing names, as they might bring in all the commonly held beliefs and associations into the data analysis. When all categories have been named, it is important to group them into higher order categories, creating subcategories that answe r when, why, where, who, what, and how questions. Identifying the characteristics of concepts (a nece ssary task to group them into abstract categories) is a challenging task in and o f itself. This is because a concept can have many apparent and less apparent characteristic s. For example, an apparent characteristic of a laptop is its ability to connec t to the Internet, and one of its less apparent characteristics is its ability to find uns ecured networks to engage in
31 unscrupulous behaviors. It is important that the c ontext in which these concepts are embedded is taken into account as the characteristi cs of concepts are identified. After categories are created, the characteristics o f categories and their dimensions must be identified. The dimension of a characteristic represents the location where the characteristic lies along a continuum. For example one characteristic of a laptop can be the frequency (or the number of times) the laptop c rashes over a given period of time, which can be dimensionalized using the word seldom. This helps differentiate these types of laptops from those that crash regularly which ultimately enable researchers to identify patterns in the data set. This in turn helps group the dat a according to these patterns and conduct a more thorough analysis. There are several ways with which open coding can b e performed. One of the most commonly used techniques, especially at the be ginning of the data analysis, is the line-by-line analysis. This approach requires anal yzing every word and phrase, and identifying relevant concepts in the data to create categories. Once categories have been generated, the researcher can use the categories to code the rest of the data. It is also possible for the researcher to analyze paragraphs o r even documents to assess similarities and differences, though line-by-line analysis is us ually more insightful. Axial coding After identifying categories, axial coding is performed to reassemble the data and develop relationships between categories and subcat egories. These relationships provide explanations about the observed phenomenon in the d ata set. Although axial coding is distinct from open coding, it can be performed simu ltaneously. Strauss and Corbin
32 suggest that there are four tasks that need to be p erformed during axial coding: (1) laying out the properties of a category and their d imensions, a task that begins during open coding, (2) identifying the variety of conditi ons, actions/interactions, and consequences associated with a phenomenon, (3) rela ting a category to its subcategories through statements denoting how they are related to each other, (4) looking for cues in the data that denote how major categories might rel ate to each other (Strauss and Corbin, 1998, p.126). The relationships between categories can be evident in the data set, rendering axial coding rather easy. However, in most cases, they can be very subtle and implicit, and require using a scheme (also referred to as pa radigm) for their identification. In doing so, researchers try to understand which categ ories represent conditions (or the circumstances in which the phenomenon is embedded), which ones represent actions/interactions (or the responses of individuals to events under t hese conditions), and which ones represent consequences (or outcomes of actions/interactions). While conditions answer the where, why, and when, questio ns; actions/interactions answer how and whom; and consequences answer questions about w hat happens as a result of the actions/interactions. As conditions, actions/interactions, and consequenc es are identified, hypotheses begin to emerge, and researchers can start explaini ng why a phenomenon occurs, under what conditions the phenomenon occurs, and what con sequences are expected when the phenomenon occurs. After hypotheses are proposed, they should be validated by identifying supporting evidence for their existence in the rest of the data. In the case of
33 contradictions, other unaccounted conditions can be sought to increase the explanatory power of the theoretical relationships. Selective coding After open and axial coding have been conducted, th e categories and the relevant relationships between them are integrated using selective coding to develop a theory. The first step of selective coding is to identify the c entral category that binds all other categories and gives them a meaning. In this sense the central category represents the main theme of the study. The central category migh t evolve from the existing categories or may be a higher order category subsuming all oth ers. Several criteria exist for testing the centrality of a category, such as being related to other categories; appearing frequently in the data; and having logical and cons istent relationships with other categories. However, having a central category does not necessa rily indicate that categories can be integrated coherently around it. The integr ation process is usually challenging and may require researchers to draw upon different tech niques such as a storyline, diagram, or memo-based approach. In the storyline approach, questions are asked about what is going on, what is the major concern here, or wh at is the data telling. Answers to these questions can pull together all the related c ategories, and thus create a cohesive story. In the diagramming technique, diagrams are used to depict relationships between categories. When all relationships are diagrammed, the diagrams are integrated with one another to reveal the central category, providing a general understanding of the phenomenon. In the memo-based technique, notes tak en during data analysis are used to
34 identify commonalities between categories and to co mbine these categories around a common theme. Once a theory is generated, it should be refined to optimize internal consistency and logic. As the first step, researchers must ens ure that the central construct has characteristics and dimensions (as described in ope n coding). If there are insufficient characteristics or dimensions, the data analysis mu st be repeated. As the second step, the researcher should ensure that the characteristics a nd dimensions of all categories show variation. For example, if frequent performers of a behavior are observed, non-frequent performers of the same behavior should also be soug ht as participants. Otherwise, additional data collection may be necessary. At th is phase, certain decisions about whether to drop certain ideas from the theory may b e necessary. It is possible that not all observations may be fully supported by the data, de spite their novelty. In such cases, these observations can be dropped from the theory t o be pursued in a future project. Finally, the theory must be validated by comparing it to the raw data. This step can be performed by researchers themselves or by an outsid er. Data collection The data collection for this study was performed in two phases. The first phase surveyed participants using face-to-face and phone interviews in addition to an online questionnaire. All data collection instruments ask ed participants whether they thought expertand/or community-governance improved knowle dge quality in electronic repositories, and why. The questions were designed to uncover the aspects of each governance mechanism that contributed to knowledge quality. The second phase of data collection sought to quantify the quality implicati ons of both governance mechanisms,
35 and using an online questionnaire asked participant s to rate the degree to which they thought expertand community-governance increased knowledge quality in the repositories used in their organizations on a fivepoint scale. The face-to-face and phone interviews used during t he first phase were semistructured, and responses were either recorded on t ape or summarized as notes during interviews. The online questionnaires that were em ployed in both phases of data collection were administered through the services o f a popular vendor on the Web using the template questions provided by the vendor. Que stions were open-ended and included comment boxes for participants to type their answer s. The questionnaires were hosted on the vendors Web servers, and were accessible using the Web link provided by the vendor. The first pages of both questionnaires pro vided instructions for participants, and briefly described expertand community-governance. The following page required participants to select the governance mechanism(s) used in their organizations. Possible answers were only expert-governance, only commun ity-governance, and both mechanisms. Depending on their answers, questions that were relevant to the chosen mechanism were presented to participants. The data collection instruments used in this study involved questions other than the quality imp lications of expertand communitygovernance. The responses related to quality outco mes are discussed here since they directly address the research question. Sample characteristics Participants in the first phase Two different groups of individuals took part in th e first phase. The first group consisted of 30 working professionals enrolled in t he Executive-MBA program of a
36 major university located in the southeastern United States. Participation in the study was part of a class activity for one of the courses in the program. Although participants responses were collected using an online questionna ire, face-to-face interviews were also conducted with five of the participants to further clarify some of the responses and preliminary findings. The second group consisted of four knowledge manage ment professionals responsible for overseeing the use of expertand/o r community-governed repositories used in their firms. These individuals were member s of a knowledge management mailing list and volunteered to be interviewed from a total of approximately 200 members. All four interviews were semi-structured and were conducted on the phone. In total, 34 professionals from 27 different firms were interviewed in the first phase of the study. Twenty-two of these (65%) iden tified themselves as managers in their current organizations, while the remaining 12 (35%) worked at senior level positions. Four of the participants (12%) were res ponsible for managing the knowledge repositories used in their organizations. The prof essionals had an average work experience of 15 years. The most senior profession al had a total of 35 years work experience, while the most junior professional had four. Twenty-nine (85%) of the participants used knowledg e repositories in their firm or organizational unit. Of these, 15 (52%) used on ly expert-governance; four (14%) used only community-governance; six (21%) used both expe rtand community-governance; and four (or 14%) did not use either of the two gov ernance mechanisms. These figures are summarized in Table 4.
37 The majority of the participants actively used know ledge from repositories in their organizations. The average frequency of know ledge use was 2-4 times a month for both expertand community-governed repositories. Although several participants mentioned that they used knowledge from repositorie s on a need basis, three consulted the repository used in their firms more than once e very day. Participants who used: Number: Knowledge repository 29 Only expert-governed repository 15 Only community-governed repository 4 Both expertand community-governed repository 6 Repository without a governance mechanism 4 No knowledge repository 5 Total 34 Table 4. Breakdown of participants in the first pha se Participants also actively provided contributions t o the knowledge repositories used in their organizations. Only two of the parti cipants never provided contributions, while six participants provided 10 or more contribu tions. Although participants provided, on average, 2-4 contributions per month, most contributions were made on a need basis. Participants in the second phase The second phase of the study was conducted using a n online questionnaire. The goal was to reach to a wider audience and determine how knowledge users rated the quality implications of expertand community-gover nance. The link to the questionnaire
38 was distributed to employees of three auditing firm s and to members of two online mailing lists. One of the mailing lists concerned general accounting principles, while the other involved enterprise resource planning (ERP) i mplementations. The response rate for the second phase of the study is estimated to be less than 1% since only 62 individuals responded to the question naire. The major reason for the low response rate was the lack of incentive. Of 62 par ticipants, only 36 provided useful responses. Among the remaining 26 participants, 15 exited the survey prematurely (after answering the first few questions), and 11 indicate d that they used neither expertnor community-governed repositories in their organizati ons. The usable data set for the second phase included r esponses from 10 different industries: information technology (IT), banking, shipping, airline, healthcare, manufacturing, audit and consulting, telecommunicat ions, insurance, and fast moving consumer goods. Forty-four percent of participants (16 out of 36) identified themselves as managers or directors in their respective organi zations. The average work experience of participants in their current position was close to five years. Participants total fulltime work experience was between 15 and 20 years. The most experienced individual had more than 20 years of full-time work experience whereas the least experienced individual had been working full-time for at least a year in their organization. The related distributions of participants work experience are presented in Figure 2.
39 Figure 2. Second-phase participants work experienc e Sixty-one percent of the participants (22 out of 36 ) used both expertand community-governed repositories in their organizati ons. Among those remaining, the number of participants who used only expert-governa nce (19.5% or 7 out of 36) was equal to the number of participants who used only c ommunity-governance (19.5% or 7 out of 36). Participants who used both governance mechanisms me ntioned that communitygoverned repositories were relatively new in their organizations compared to expertgoverned repositories. For example, one participan t had been using an expert-governed repository for more than five years, but a communit y-governed repository for only three years. However, community-governed repositories el icited more contributions relative to expert-governed repositories. On average, particip ants made 2-4 contributions to expertgoverned repositories per month, and 2-4 contributions to community-governed repositories per week The characteristics of the participants who used on ly expertand only community-governed repositories were also similar t o those who used both. In the case of only expert-governance, a typical participant ha d used the repository for nearly three 0 2 4 6 8 10 12 Less than 1 year 1-2 years 3-5 years 6-10 years 10+ years Experience in current position 0 5 10 15 20 Less than 1 year 1-2 years 3-5 years 6-10 years 11-20 years 20+ years Total full-time work experience
40 years, whereas in the case of only community-govern ance, they had used the repository for nearly two years. Data Analysis Data analysis was performed by the researcher. Aft er data collection was over, the tape-recorded interviews and handwritten notes were transcribed into an electronic format, and the responses to the online questionnai res were downloaded. The combined data archive was analyzed using the coding techniqu es described earlier. In order to demonstrate the data analysis process, coding of one of the factors that contributed to knowledge quality in expert-governed repositories is described below. In the first step of coding, comments related to why s ubjects thought expert-governance improved knowledge quality were identified from the data set. The majority of the comments were obtained from the online questionnair e. These comments were short statements typed into comment boxes provided for th e related question in the online questionnaire. Example statements for expert-gover nance are presented in Table 5. In the second step, open coding was performed, in w hich comments, such as those presented in Table 5, were scrutinized line-by-line to identify candidate concepts that articulated participants beliefs about expert-gove rnance and knowledge quality. For example, the first comment in Table 5 shows three c oncepts identified using in vivo codes as highlighted in the original response: gatekeeping evaluating and correcting Similarly, in the second comment, the participant m entioned that high quality knowledge in the expert-governed repository was achieved thro ugh reviewing scrubbing editing and reduction (as highlighted in the original text).
41 Participant comment Concepts Category Experts are like gatekeepers. They evaluate, corre ct and post [documents] to the [repository] and give acces s to all stake holders. So the quality is never compromised Gatekeeping, evaluating, correcting Governance functions [Content in expert-governed repository is] very hig h quality. It's all been through multiple reviews, a nd scrubbing, and editorial work, and reduction. Ther e isn't anything in there that hasn't been looked over thre e or four times... Seriously... Reviewing, scrubbing, editing, reduction [Expert-governance] makes sure that no false inform ation is deliberately inserted in the knowledge repositor y and misleads users. Filtering [Expert] vetting helped in identifying the appropri ate online site faster. Vetting Table 5. Participant comments for quality implicati ons of expert-governance Following the identification of concepts, similarit ies and differences between these concepts were examined to create higher order categories (hereafter referred to as factors ). For example, the similarity between the concept s identified in Table 5 was that they described actions or interventions performed b y experts to address knowledge quality. Therefore, these concepts were grouped to gether, creating the first factor that contributed to knowledge quality in expert-governed repositories, namely governance functions Using the same technique for the rest of the commen ts identified two more factors: credibility of experts and ownership of content The concepts that guided the identification of these two factors are presented i n Table 6. The table shows that some concepts can be considered factors without being gr ouped with other similar concepts. This occurred because the identification of concept s and factors were performed simultaneously instead of sequentially as suggested by Strauss and Corbin (1998). For example, once the ownership concept was identified in one of the comments prov ided for
42 expert-governance, it was used as a higher order fa ctor to code the rest of comments that tapped into the same concept. Following open coding, axial coding was performed t o identify relationships among factors, building further understanding about paradigm model proposed by Strauss and Corbin (1998). Strauss and Corbin sugg est that during axial coding researchers should define such a model that consist s of actions conditions and consequences in order to identify which factors are the most sa lient. The model builds on the position that actions and conditions make up th e ingredients for consequences, and thereby, help researchers develop hypotheses about the observed phenomenon. Expert-governance Concepts Factor Gatekeeping, evaluating, correcting, vetting, filtering, reviewing, scrubbing, editing, reduction Governance functions Ownership Ownership Expertise, knowledge, trustworthiness, reliability Credibility Table 6. Concepts and categories identified for exp ert-governance In the context of this study, the consequence aspec t of the paradigm model was knowledge quality in electronic repositories, and was set a priori during data collection. The question that was used in interviews and the on line questionnaire was the research question guiding this essay, which asked participan ts whether they thought expertand community-governance improved knowledge quality, an d why or why not. The phrase because was implicit in all responses, which esta blished an axial relationship between the three factors identified during open coding and the category of interest to this study,
which is knowledge quality governance functions was contents were the conditions relationship presented in actions and conditions, and explained a higher order Figure 3 Hierarchical structure of It is important to note that the questions used in interviews and the online questionnaire were targeted and directly addressed the research question of this essay. The central factor (knowledge quality) was set duri ng axial coding rather than during sele ctive coding. Strauss and Corbin as the last step of the coding process, which helps develop a unifying story around a central factor (or construct) to addres essay knowledge quality the central factor and the other factors) provided a full and plausible explanation as to Governance functions (Action) 43 knowledge quality For this reason, knowledge quality was the was the action and the credibility and the ownership of conditions of the paradigm model. This suggested a hierarchical presented in Figure 3 in which the three sub-categories represented and explained a higher order factor, namely knowledge quality Hierarchical structure of constructs for expertgovernance It is important to note that the questions used in interviews and the online questionnaire were targeted and directly addressed the research question of this essay. The central factor (knowledge quality) was set duri ng axial coding rather than during ctive coding. Strauss and Corbin (1998) suggest that researchers use selective coding as the last step of the coding process, which helps develop a unifying story around a central factor (or construct) to addres s the research question. The central factor in thi s knowledge quality and the relationships identified during axial codin g (between the central factor and the other factors) provided a full and plausible explanation as to Knowledge quality Governance functions Credibility Ownership of content (Action) (Conditions) (Consequence) the consequence ownership of a hierarchical represented the knowledge quality governance It is important to note that the questions used in interviews and the online questionnaire were targeted and directly addressed the research question of this essay. The central factor (knowledge quality) was set duri ng axial coding rather than during suggest that researchers use selective coding as the last step of the coding process, which helps develop a unifying story around a s the research question. The central factor in thi s and the relationships identified during axial codin g (between the central factor and the other factors) provided a full and plausible explanation as to
44 how expert-governance affected knowledge quality. For this reason, selective coding and axial coding were completed simultaneously. It should be noted that the coding process explaine d above was also used for community-governance as a means to assess participa nts interpretation of the effects of community-governance on the quality of knowledge in electronic repositories. The next section summarizes these findings for expert-govern ance and discusses the findings for community-governance further. Findings Factors that contribute to knowledge quality The research question of interest was whether exper tand community-governance improved knowledge quality in organizational reposi tories, and why or why not. The data revealed that both governance mechanisms impro ved quality of knowledge in repositories. Especially in the second phase of th e study, when participants were asked to rate the governance mechanisms according to the ext ent to which they improved knowledge quality, participants rated expert-govern ance with a score of 4.2 (based on a five-point scale; 1 being not at all and 5 being to a great extent), and communitygovernance with a score of 4.4 (based on the same f ive-point scale). Although the difference between the two scores was not significa nt statistically, the fact that participants rated both mechanisms high on the scal e provides evidence for the efficacy of both governance mechanisms in increasing knowled ge quality. In order to address the why part of the research question, participants comments were analyzed using the coding procedure e xplained in the data analysis section. In the case of expert-governance, the ana lysis revealed that three different
45 factors contributed to knowledge quality in electro nic repositories: (1) governance functions employed by experts, (2) experts credibility and (3) experts ownership of content published in repositories. The first factor governance functions represent s actions, such as gatekeeping, evaluating, correcting, vetting, filtering, reviewi ng, scrubbing, editing, and reduction that are performed by experts to increase knowledge qual ity. The relationship between governance functions and knowledge quality is an ex pected finding. Since governance functions are central to any implementation of expe rt-governance, it is intuitive for individuals to associate the execution of these fun ctions with higher quality knowledge. However, the execution of governance functions alon e may not be sufficient for higher quality. For instance, one participant observed th at the way these functions are executed may also play a role in improving knowledge quality : [Content in expert-governed repository is] very hi gh quality. It's all been through multiple reviews, and scrubbing, a nd editorial work, and reduction. There isn't anything in there that hasn't been looked over three or four times ... Seriously... (emphasis added). This suggests that governance functions were iterat ive repeated several times before submissions were published in the repository Although this may suggest that the number of times the governance functions are execut ed may matter (and a higher number of iterations resulting in higher knowledge quality ), the participants comment connotes thoroughness rather than the literal number of occurrence. Thi s is because each time a governance function is repeated, it adds to the ove rall knowledge quality by addressing the issues that had been overlooked previously. Th is, in turn, implies that the thoroughness of execution matters more than the number of times the governan ce functions are executed. Even if governance functio ns are executed numerous times, they
46 may not contribute much to knowledge quality if the y are not executed thoroughly. This view was corroborated by another participant a se nior executive in the IT industry who was responsible for overseeing the expert-gover ned repository. The participant considered the experts workload a serious impedime nt to achieving high quality knowledge in the repository, because experts were n ot able to vet the submissions made to the repository thoroughly When these individuals were expected to vet all submissions in addition to performing their day-today tasks, this produced a major bottleneck in the development of the knowledge base of the firm. It usually took several months for the experts to execute the governance fu nctions after contributions were submitted to the repository. Though not advised by their supervisors, these individuals traded off the thoroughness of the vetting processe s for a higher throughput. They started to vet the contributions quickly, which posed a thr eat to the overall quality of these contributions. The second aspect of expert-governance that emerged from the data as a contributor of knowledge quality was the experts credibility Prior research conceptualizes credibility using four dimensions: k nowledge, trustworthiness, expertise, and reliability of individuals (e.g., Sussman and S iegal, 2003). Participants responses about the quality implications of expert-governance tapped into these dimensions, indicating that credibility of experts was a signif icant criterion related to the quality of knowledge in expert-governed repositories. One par ticipant commented, [Content in expert-governed repository is of high quality], because it is completed by the experts in that subject matter. However, these people don't always use this informa tion on a daily basis like others. (emphasis added)
47 The word experts is used in the context of subjec t matter expertise the extent of experts knowledge of the domain of interest. T his highlights the centrality of the contribution of individuals knowledgeable in their domains to the quality of knowledge in repositories. Another participant highlighted the reliability aspect of experts, There is credibility to [expert-governed repositor ies]. You do not have the distrust and risk of incorrect information Expert[s] tend to [weigh] everything from all angles and they are pretty reliable . (emphasis added) Others associated high quality knowledge with the t rustworthiness of experts, [The expert-governed repository] provides informat ion by known and trustworthy experts who have long [years of] experience in the field. The experts ensure that everything stored i n [the repository] is [of] high quality. (emphasis added) All the above comments emphasize the contributions of knowledge, reliability, and trustworthiness of experts to the quality of kn owledge in repositories. Following the procedures for selective coding (Strauss and Corbin 1998), these concepts were combined to a higher order factor, namely the credibility of experts who perform the governance functions. Credibility, by nature, vari es along a high-low dimension. The comments presented above, fall toward the high en d of the spectrum, suggesting that the quality of knowledge in expert-governed reposit ories is directly related to the credibility of experts. The empirical data gathere d were elicited using questions to stimulate consideration of factors that are positiv ely related to knowledge quality. Consequently, few comments relate to the absence of credibility: nevertheless, the contrary should also hold, where content governed b y less credible experts would be perceived as being lower in quality. The last aspect of expert-governance that was ident ified in the data as a contributor of knowledge quality was experts ownership of content stored in
48 repositories. It is important to note that there a re at least two types of content ownership in the context of this study: (1) ownership as a re sult of individuals associating or identifying themselves with contents, and (2) owner ship as a result of content authorship. This study suggests that the first type of ownershi p is salient to expert-governance and knowledge quality, because the comments provided by participants connote experts identifying themselves with the content rather than authorship. For example, The gatekeepers should have pride and ownership o f the contents which [mean] higher quality contents. Community-gov ernance may have tragedy of the commons syndrome, to put it in very simplistic term[s]. Similar to the experts credibility, experts owner ship of content is also a condition that affects knowledge quality in reposit ories. Further, ownership varies along a high-low dimension, indicating that experts with high a strong sense of ownership contribute substantially more to the quality of kno wledge. It is noteworthy that experts can have feelings of ownership toward either contri butions or repositories. In the former case, experts can have feelings of ownership only t oward those contributions that are vetted by themselves. In this case, experts may no t care much about contributions vetted by other experts. In the latter case, experts can have feelings of ownership toward the entire repository regardless of the extent of contr ibutions they vetted: experts may be more vigilant about all contributions and feel resp onsible for the overall quality of the repositories. In summary, three factors were mentioned by partici pants as being salient for improving knowledge quality in electronic repositor ies: (1) thorough execution of governance functions, (2) credibility of experts, a nd (3) experts ownership of contents published in repositories. Three propositions are advanced from this analysis:
49 P1a: Thorough execution of governance functions is positively associated with high quality content in expert-gove rned repositories. P1b: Credibility of individuals, who perform the go vernance functions, is positively associated with high quali ty content in expert-governed repositories. P1c: Experts ownership of published content is pos itively associated with high quality content in expert-gove rned repositories. In the case of community-governance, the coding pro cess identified two factors that contributed to knowledge quality: (1) governance functions employed by community members, and (2) communitys involvement in the governance process. The concepts that make up these factors are presented i n Table 7. Community-governance Concepts Factors Multiple edits, editing, rating, reviewing Governance functions Seeking opportunities, taking action, involvement, taking initiative Involvement Table 7. Concepts and categories identified for com munity-governance As was found from the data exploring expert-governa nce, participants identified governance functions as a factor affecting the quality of knowledge in communitygoverned repositories. The governance functions re presented different types of actions such as editing, reviewing, and rating performed by community members. For example, The information was extremely well-organized and e asy to peruse. It also had many of the examples I was loo king for. If [this information] wasnt edited by multiple indivi duals, it wouldnt be this valuable for me.
50 This highlights the importance of multiple edits an d suggests that edits provided by community members affected the organization and readability of the knowledge asset. Further, edits contributed to knowledge quality thr ough the provision of relevant examples. The immediacy of the value perceived by this participant suggests a substantial contribution to the quality of the know ledge through editing. Another participant mentioned the importance of editing, re viewing, and rating for achieving high quality knowledge, Developers and managers [do not] always remember e very single detail on every single project; full-fledged commun ity governance not [only] enables the users to share content, but also serves as [a] valuable knowledge base which can be continuously i mproved upon by [its] members through editing, rating, and review activities. The salience of governance functions in improving k nowledge quality in community-governed repositories is an expected find ing. Unless members of the community execute governance functions, it is not p ossible to improve or signal knowledge quality in community-governed repositorie s. Unlike expert-governance, the comments in the data set do not provide evidence ab out the thoroughness of governance functions. Instead, the comments suggest that gove rnance functions may vary along a diversity dimension, indicating that the range of community members involved in executing the governance functions may affect knowl edge quality. For example, in the first comment, the phrase multiple edits suggests that the knowledge asset was edited by different individuals, all of whom provided diff erent insights collectively. Therefore, quality improvement was not achieved using a single revision cycle (typical of expertgovernance), but through the collective effort of i ndividuals. This is similar to the notion of the wisdom of crowds (Surowiecki, 2004), which s uggests that the aggregate
51 information possessed by the individuals in a group is always superior to the information possessed by a single individual in that group. Th erefore, it is reasonable to argue that the execution of governance functions by different members in the community improves the quality of a knowledge asset more than the exec ution of governance functions by a single member in the community (as in expert-govern ance). This suggests that the diversity of members who execute the governance fun ctions is a salient dimension of governance functions for achieving high quality kno wledge. It is also noteworthy that governance functions in community-governance can increase knowledge quality continuously (as mentioned by the second participant above), unlike expert-governance. This is an interesting f inding, as it highlights one structural difference between expertand community-governance described earlier in this essay. As conceptualized in this study, community -governa nce is a post-publication process and it allows the quality of a knowledge asset to b e improved during its lifetime or during the lifetime of the repository. Further, it does n ot impose any restrictions on community members to execute governance functions. Therefore as long as contributions are accessible in the repository, community members hav e the opportunity to make modifications or provide suggestions, increasing th eir quality. This contrasts with expertgovernance a pre-publication process which does not allow further improvements to be made to contributions (unless organizational mem bers make formal change requests to experts, who then contract out the modification eit her to the original contributor, or to another organizational member). Further, expert-go vernance restricts user-privileges and lets organizational members use knowledge assets on ly without providing any feedback in return. This, in turn, may cause knowledge asse ts to become outdated very quickly,
52 unless the original contributor (or a current user) of that knowledge asset file a modification request to experts. This issue was co rroborated by one participant in the IT industry who was responsible for overseeing both th e expertand the communitygoverned repositories in his firm. The participant suggested that content in the expertgoverned repository was more prone to becoming outd ated than content in communitygoverned repository, since it did not allow anybody (other than experts) to edit those contributions. The second aspect of community-governance that cont ributed to knowledge quality was the involvement of community members in the governance process. T he related concepts identified in the data involved se eking opportunities for enhancing quality, taking initiative, taking action, and bein g involved. One participant, who was using a community-governed repository in the teleco mmunications sector said, When enough eyes look at a single document, its qu ality inevitable increases of course if people take act ion for improving quality. But I think the [communitys] involveme nt also matters. If [community members] do not take initiative whi ch is sometimes the case in our company don't expect to have qual ity information regardless of how many people look at it. The data also provided evidence for the effect of l ack of involvement on knowledge quality. In this case, lack of involveme nt was mentioned as a major drawback of community-governance in improving quality. One participant mentioned that the knowledge quality in the community-governed reposit ory (i.e., the wiki) used in the company did not provide high quality content, becau se, People rarely edit the wiki content, because they dont think this is expected of them. Whereas experts roles and responsibilities are for mally defined in expertgovernance, such formalization is lacking in commun ity-governance. Unless community
53 members are formally assigned the governance functi on, community-governance may not affect knowledge quality. There are many reasons w hy community-members may not get involved in the governance process. One might be t he lack of incentives to govern knowledge assets. Several interviewees mentioned t hat their organizations did not reward contributions made to community-governed rep ositories (such as wikis or discussion forums), let alone efforts to assess and improve the quality of contributions stored in these repositories. Therefore, in the ab sence of adequate incentives, community members are unlikely to spend their valuable resour ces (such as time and cognitive effort) in governing knowledge assets. In summary, the data suggest that two aspects of co mmunity-governance contribute to knowledge quality in repositories: (1 ) executing the governance functions continuously and by a diverse group of members, and (2) the involvement of community members in the governance process. Two proposition s are advanced from this analysis: P2a: Executing governance functions continuously an d by a diverse set of individuals is positively associated with high quality content in community-governed repositories. P2b: Community members involvement in governance i s positively associated with high quality content in expert-governed repositories. The discussion above addresses the research questio n of this study. However, the data revealed two other interesting insights worthy of discussion about expertand community-governance. The first of these concerns users perceptions of expertgovernance. Participants in this study associated expert-governance with accreditation and stated that the involvement of experts during t he knowledge transfer process provided them with additional assurance about the q uality of knowledge stored in repositories. One interviewee said,
54 when [information] comes from [the expert-governed repository] it makes a lot of difference, because [experts] hav e thought through this and seen it from every aspect and angl es. Its pretty much a complete and correct solution. In a way, involvement of experts positively biased users perceptions of knowledge stored in expert-governed repositories. A participant commented, [experts] lend credibility to the material and mak e it more meaningful than if just anybody published the infor mation. This comment is particularly interesting, because i t indicates that individuals may have more favorable attitudes toward an expert-gove rned knowledge asset even if its quality does not significantly differ from the qual ity of a community-governed (or even an ungoverned) knowledge asset. Individuals tende ncy to perceive expert-governed knowledge assets as more meaningful (or of being hi gher quality) may prevail even if they are unaware of the quality control processes o r experts level of expertise. This view is borne out by a participant who said, I have more confidence in the information knowing that it was vetted by experts compared to wikis. I know (hope) the experts know their subject. Although several participants perceived expert-gove rnance as an accreditation process, there were others who were skeptical of th is so-called accredited knowledge: one interviewee commented, I believe it is still important to be critical of the information, but it is a lot more reliable than the Internet. Another interviewee said, You should always [check] the accuracy and validit y of information presented to you to some degree The second additional insight gained from the data analysis concerned the implications of community-governance on social rela tions in organizations. Several
55 participants mentioned that community-governance ha d built a collaborative environment in their organizations, and induced greater level s of interaction among employees. One interviewee mentioned, [Community-governance] not only enables us to shar e content, but also serves as a valuable tool for interaction The socialization and collaboration enabled by comm unity-governance also transcended the electronic medium. One interviewee in the IT industry stated that community-governed repositories fostered interactio ns among employees not only through electronic repositories, but also through f ace-to-face discussions. The interviewee explained that he engaged in several fa ce-to-face and phone discussions with colleagues, after he provided a comment about a com mon software problem discussed in the community-governed repository of the firm. If the repository were expert-governed and did not enable individuals to communicate their ideas online, the participant would not have engaged in face-to-face or phone discussio ns. The additional insights gained from the interview d ata show that, first, participants perceive expert-governance as an accre ditation process (despite the skepticism of certain participants), and second, co mmunity-governance foster a more collaborative environment. Assessment of knowledge quality Although the above analysis and discussion focuses on knowledge quality as the dependent variable of interest, it does not directl y address participants perceptions of knowledge quality. Therefore, this section present s the findings about how individuals assessed the quality of knowledge they used from th eir organizational repositories. For this purpose, participants responses to one of the questions used in the online
56 questionnaire were used, which asked participants t o recall the last piece of knowledge they used from their organizational repository and explain how they assessed its quality. Table 8 summarizes participants perceptions of qua lity. The coding techniques described earlier were used to develop the factors in the table. Participants assessed quality based on two aspects of knowledge, its appl ication in a given context and its goodness. The application of knowledge concerned whether using the knowledge in a given context led to successful outcomes, advised an efficient solution, and fit the problem at hand. Assessments based on the goodness of knowledge involved a number of characteristics of the contribution retrieved fr om the repository such as readability, precision, sufficiency, accuracy, timeliness, and a ccessibility. Concepts Factors Higher order factors Working solution, successful application, resolve the problem, usefulness Successful application Application of knowledge Efficient solution, time it takes to apply Efficiency of solution Customized solution, fit to actual process Fit to situation Easy to follow, well-organized, easy to peruse Readability Goodness of knowledge To the point, precise Precision Sufficient information, existence of examples Sufficiency Correct Accuracy Up-to-date Timeliness Easy access Accessibility Table 8. Concepts and factors identified for qualit y The two criteria used for assessing knowledge quali ty differ in two respects. First, assessments made using the application of kn owledge are more contextual, as the
57 context in which knowledge is applied plays a role in determining the quality of the knowledge. In comparison, assessments made using t he goodness of knowledge is context independent, as participants make evaluatio ns based on its general characteristics that are not bound by the context. Second, assessm ents made using the goodness of knowledge can be made before knowledge is actually applied, whereas assessments about the application of knowledge can be made only after knowledge is actually applied. It is interesting to note that some of the concepts and factors presented in Table 8 tap into the dimensions of data and information qua lity in the extant literature. The categories identified for goodness of knowledge (i. e., readability, precision, sufficiency, accuracy, timeliness, and accessibility) were the s ame as some attributes of data and information quality suggested by prior studies. Re search on data and information quality has a long history and researchers have been trying to define data and information quality for a long time. One of the most cited works is Wa ng and Strong (1996), who organize the attributes of data quality (DQ) into four dimen sions: intrinsic, contextual, representational, and accessibility. They suggest that Intrinsic DQ denotes that data have quality in their own right. Contextual DQ highlights the requirement that data quality must be considered within the context of th e task at hand. Representational DQ and accessibility DQ emphasize the importance of the role of systems (W ang and Strong, 1996, p.6). The attributes identified for each of these dimensions are presented in Table 9. It is important to note that the attribut es identified by Wang and Strong (1996) apply not only to data, but to processed data (or i nformation) as well. Similarly, Zmuds (1978) quality attributes for hardcopy reports, and Goodhues (1995) quality attributes for
58 patient records show that attributes of data qualit y extend to information quality as well. These quality attributes are summarized in Table 9 adapted from Lee et al.(2002). Study Intrinsic Contextual Representational Accessi bility Wang and Strong (1996) Accuracy, believability, reputation, objectivity Value-added, relevance, completeness, timeliness, appropriate amount Understandability, interpretability, concise representation, consistent representation Accessibility, ease of operations, security Zmud (1978) Accurate, factual Quantity, reliable/timely Arrangement, readable, reasonable Jarke and Vassiliou (1997) Believability, accuracy, credibility, consistency, completeness Relevance, usage, timeliness, source currency, data warehouse currency, non-volatility Interpretability, syntax, version control, semantics, aliases, origin Accessibility, system availability, transaction availability, privileges Delone and McLean (1992) Accuracy, precision, reliability, freedom from bias Importance, relevance, usefulness, informativeness, content, sufficiency, completeness, currency, timeliness Understandability, readability, clarity, format, appearance, conciseness, uniqueness, comparability Usableness, quantitativeness, convenience of access Goodhue (1995) Accuracy, reliability Currency, level of detail Compatibility, meaning, presentation, lack of confusion Accessibility, assistance, ease of use (of hardware, software, locatability Table 9. Dimensions of knowledge quality identified in the literature An interesting finding of this study is that the fa ctors identified for goodness of knowledge tapped into all four of the dimensions of knowledge quality presented in Table 9, whereas the other factors identified for a pplication of knowledge do not map to these dimensions: they are largely missing in the e xtant literature. This can be attributed to the distinction between data and information and knowledge, and the different criteria
59 that are used to assess the quality of each. As me ntioned earlier, data comprise raw facts, information is processed data, and knowledge is inf ormation that has a context and that is given interpretation and meaning. It has been ackn owledged that it is difficult to make clear cut distinctions between data, information, a nd knowledge (Davenport, 1997). However, most studies agree that data, information, and knowledge can be considered a hierarchy, data being at the bottom, and knowledge being at the top. The findings of this study suggest that while the existing dimensions of quality may be valid for the entire hierarchy as a whole, we may need new dimensions of quality as we move up the hierarchy due to the differences between the two ex tremes. One such dimension may be the application of knowledge as reported in this st udy. Trustworthiness of findings A major concern of researchers using qualitative an alysis and an interpretive paradigm is the trustworthiness of findings. Since the criteria used by the positivist paradigm are not relevant to the interpretive parad igm, new approaches to judging the trustworthiness of findings have been proposed. Li ncoln and Guba (1985) suggest that four criteria, adapted from the positivist paradigm can be used to judge the merits of qualitative research: credibility, transferability, dependability, and confirmability of findings. Credibility taps into the internal validity criteri on of the positivist paradigm, and assesses whether or not the study is an accurate re presentation of the reality being investigated. In order to ensure credibility, rese archers can take several precautions, one of which is to stay in the field for a sufficiently long time to engage with a number of cases. The goal is to make sure that researchers l earn as much as possible from the field
60 about the topic of interest. Another precaution is the use of triangulation, which requires researchers to use multiple sources for data collec tion. Through the use of triangulation, researchers may collect data from interviews, obser vations, focus groups, archival data, and any other supporting documents. Triangulation can also be achieved by interviewing people from different parts of the organization, di fferent departments, or hierarchical levels. A third precaution is taking negative case s into consideration during data collection as well as positive ones. This not only ensures that there is variation in the data set (especially in the dependent variable), bu t also increases the explanatory power of the theory by reconciling the differences betwee n positive and negative cases. A fourth precaution involves discussing the ideas and findings obtained from the data with peers and senior researchers. In this way, researc hers can exchange ideas with other researchers or even with practitioners to determine whether the data analysis lends itself to alternative interpretations. In order to ensure the credibility of this study, s everal actions were taken during the course of the investigation. First, data colle ction was performed in two different phases from various organizations in different indu stries to increase the likelihood that the responses consistently construct the reality as closely as possible to the natural setting. Second, several in-depth interviews were conducted with practitioners to uncover as much as possible about expertand commu nity-governance, and to determine whether there were alternative explanations for the findings. Two of the face-to-face interviews were conducted after the initial phase o f data collection, providing an opportunity to discuss the preliminary findings wit h experienced practitioners in the field. Both interviewees agreed that the findings were not only highly representative, but also
61 fully comprehensive of the quality implications of expertand community-governed repositories in their organizations. As the third step, the research methods, the data collection techniques, and the preliminary findings were discussed with dissertation committee members and presented at a research sympo sium. These discussions ensured that the processes used in the study were capable o f constructing the reality adequately. The second criterion, transferability, relates to t he external validity (or generalizability) aspect in the positivist paradigm and involves the applicability of findings in other contexts or to other populations. This is one of the major concerns of qualitative research, since findings are usually ba sed on a small number of observations. However, Lincoln and Guba (1985) suggest that resea rchers are not capable of making this judgment, as they may not know upfront what ty pes of contexts the readers may want to generalize the findings to. The most appropriat e precaution is for the investigator is to provide as much contextual information as possible, so that readers themselves can decide whether the findings can be transferred to a context of interest. In doing so, researchers can provide descriptive statistics abou t cases, the case selection criteria, data collection procedures, and other contextual data re levant to the research environment. The transferability criterion was addressed in this study by providing details about the sample selection criteria, the descriptive stat istics of participants, and other contextual details whenever direct quotes or anecdotes were us ed from participants. The fact that data were collected from a variety of individuals i n a range of organizations in various industries further enhanced the potential transfera bility of the findings, since the research used a heterogeneous sample rather than a more homo geneous one (more usually found in case studies).
62 The third criterion, dependability, taps into the r eliability aspect of the positivist paradigm, and concerns the repeatability of the fin dings. It suggests that if the same study is conducted in the same context using the sa me sample with the same data collection technique, the same findings should be o btained. In order to ensure dependability, an internal audit can be conducted t o check whether the study conforms to accepted research standards. Further, researchers can report the processes used for data collection and data analysis in detail not only to show that proper research practices were followed, but also to demonstrate that the same fin dings should be observed if the same processes are repeated. The dependability criterion of this study was addre ssed by providing details in the research methods section about the processes used f or data collection and data analysis. Further, the research practices used in this study were vetted by the dissertation committee and other experienced researchers, which ensured that appropriate techniques were used to collect and analyze the data. Althoug h the dissertation committee may not substitute an internal audit, it ensures that the s tudy conformed to standard academic practices in the field of management information sy stems. The fourth and final trustworthiness criterion is c onfirmability, which addresses the objectivity aspect of the positivist paradigm. Confirmability ensures that findings are based on the experiences of individuals (or cases) rather than the preferences or perceptions of researchers. In order to optimize c onfirmability, researchers can use the triangulation technique discussed earlier. Multipl e sources of information reduce the tendency for researchers to bias the data analysis. Besides triangulation, researchers should also accurately record each interview, take careful notes during observations, and
63 should employ good data management practices to min imize bias. As a final precaution, the processes used for data collection and data ana lysis can be audited by peers or senior researchers to ensure that findings are reported fr ee of the researchers preconceptions or convictions. The confirmability of this study was mainly satisfi ed by the data collection technique employed for this study. The majority of the interviews were conducted online, which required interviewees to type their a nswers into comment boxes provided for each question. This ensured that the responses were recorded accurately by interviewees, and were not affected by the research ers subjective understanding. Further steps taken to ensure confirmability were the detai led presentation of the data analysis process in the research methods section, and the in volvement of the dissertation committee in auditing the research practices perfor med during data analysis. Discussion Key findings The goal of this essay was to set the conceptual fo undations of knowledge governance in electronic repositories, and examine the aspects of expertand communitygovernance that contributed to knowledge quality. Following a review of the basic concepts underpinning KM, this essay surveyed the s ocietal governance literature, and extended the mechanisms associated with the governa nce of societies to the KM context to increase the understanding of the different type s of mechanisms affecting the quality of knowledge in repositories. Specifically, four diff erent governance mechanisms were identified, and two expertand community-governa nce were discussed in detail due to their popularity and prevalence in organizationa l settings.
64 Expert-governance is a centralized mechanism, where a designated group of experts act as gatekeepers to increase knowledge qu ality in repositories. Being a prepublication process, expert-governance requires eac h contribution made to the repository to be vetted by experts before publication. The ve tting process includes various tasks, some of which include evaluating contributions to c heck their accuracy, and correcting, formatting, scrubbing, editing, indexing, categoriz ing, or requesting additional information from the contributor. Some of these ta sks need not necessarily be performed by the expert, but by the contributor of the inform ation through several rounds of revision. This essay also defined community-governance as a d ecentralized mechanism, where a community of individuals affect contributio n quality in organizational repositories collectively. In this essay, communit y represents a group of individuals who share the same job description, who work in the sam e domain, or who are part of the same business process in the same organization. Co mmunity-governance is a postpublication process, as members of the community af fect the quality of contributions that have already been published in organizational repos itories. It enables community members to edit contributions (such as in wikis), p rovide comments (such as in discussion forums), or perform other functions such as rating for signaling quality. Lack of conceptual development in governance mechan isms prompts many research questions. As the first step of a longerterm research agenda, this study assessed whether expertand community-governance helped inc rease quality of knowledge in repositories, and to explore why or why not. Data collected from participants from a range of organizations revealed several important i nsights. First, both expertand
65 community-governance increased knowledge quality in electronic repositories. In the case of expert-governance, participants mentioned t hat three aspects of expertgovernance contributed to knowledge quality: (1) ex ecuting the governance functions thoroughly, (2) experts credibility, and (3) exper ts ownership of contents in the repository. In the case of community-governance, p articipants identified two aspects of community-governance that contributed to knowledge quality: (1) executing the governance functions frequently by a diverse group of individuals, and (2) community members involvement in governance. Besides the aspects of governance mechanisms that c ontributed to knowledge quality, the data revealed two other interesting fi ndings. First, participants associated expert-governance with accreditation, and suggested that the existence of expertgovernance provided them with assurance that the co ntents of repositories were of high quality. Second, participants indicated that commu nity-governance spurred socialization among community members, and fostered a more collab orative environment in organizations. These findings treated knowledge quality as a black box and did not address the meaning of knowledge quality for participants of th is study. Therefore, a post-hoc analysis was conducted to explore how participants assessed quality as they used knowledge from electronic repositories. The findin gs suggested that participants made quality assessments based on two high-level dimensi ons of knowledge: (1) the application of knowledge, and (2) the goodness of k nowledge. Assessments based on the application of knowledge were context-specific and were made after knowledge was applied in a context. They concerned whether the s pecific piece of knowledge used
66 successfully solved the problem, whether it offered an efficient solution, and whether it was a good fit for the problem at hand. On the oth er hand, assessments based on the goodness of knowledge were context independent and were made before knowledge was applied. These assessments were made based upon th e readability, precision, sufficiency, accuracy, timeliness, and accessibility of knowledg e assets and were in line with the assessment criteria used for data and information q uality in the extant literature. Limitations of the study The findings need to be interpreted within the limi tations of this study. First, the majority of the responses used in this study were t o online questionnaires. Although this increased the total number of professionals who par ticipated in the study, and thus allowed the investigator to tap into a wide range o f perspectives, the responses provided by these professionals were not as rich as the ones obtained from face-to-face interviews. Since typing answers into comment boxes takes more time and effort than providing verbal answers, participants experienced fatigue mu ch faster when having the online questionnaire. Therefore, the majority of the part icipants provided one to two line answers for most questions. This hindered the rese archers efforts in making more complex inferences from the data collected for this study. Further, the online questionnaire did not allow the researcher to ask f ollow-up questions or drill down from specific answers. This, in turn, limited the possi bility to develop stronger theoretical relationships for various concepts identified in th e study. For this reason, future phases of this research will put more emphasis on conducti ng face-to-face interviews, and use online questionnaires only as a means to increase t he sample size or tap into other perspectives not available through face-to-face int eractions.
67 Second, this study used a sample of convenience to investigate the research questions of interest. The participants were recru ited from the researchers professional network as opposed to using a systematic approach s uch as random sampling. The sampling frame for this study constituted the stude nts of the Executive-MBA program of a university, the members of various mailing lists, and the members of several auditing firms. The use of the convenience sample limits th e generalizability of findings to other contexts and organizations. Although the participa nts represented different industries, and thus helped the researcher tap into different p erspectives, future work will use more systematic approaches (such as random sampling) to ensure that the sample selection criteria do not bias the findings. Further, future research can employ the case research method (preferably in multiple organizations) as op posed to survey tools, enabling deeper exploration of the quality implications of governan ce mechanisms and identify other candidate aspects of governance mechanisms that wer e not identified in this study in. Third, the empirical data collected from participan ts was analyzed by the researcher. Independent coders were not used durin g open coding, which is the building block of the findings reported in the essay. This threatens the confirmability of the findings (Lincoln and Guba, 1985), since the resear chers preconceptions or convictions may have tainted the data analysis. Future researc h should use multiple and independent coders who are not familiar with the goals of the s tudy to develop a more objective set of findings and thus increase the confirmability of th e study. Fourth, the expertand community-governed reposito ries examined in this study are high-level abstractions, and may subsume differ ent types of technologies currently used in organizations. For instance, discussion fo rums and wikis are considered as
68 community-governed repositories in the context of t his study, although they exhibit different characteristics. However, the questions used for data collection (especially the ones used in the online questionnaire) did not ask participants the specific type of technology in use. This, in turn, eliminates the p ossibility to assess whether the quality implications of governance mechanisms also depend i n some way on the technological design of knowledge repositories. Further, it does not allow the researcher to make any detailed inferences about the aspects of specific t echnologies that employ communitygovernance as a means to affect knowledge quality. Therefore, future studies will determine the specific type of technology used in o rganizations for knowledge transfer (such as discussion forums, wikis, intranet pages, or file servers) before categorizing them as expertor community-governed repositories. This may help categorize the nature of the interplay between the technological f eatures and the efficacy of governance mechanisms. Theoretical implications This study has several theoretical implications. F irst, it offers propositions about the aspects of expertand community-governance tha t increase knowledge quality. Although expertand community-governance are becom ing more common in many organizations, the limited number of studies in the literature shed very little light on how these mechanisms contribute to knowledge quality. The propositions offered in this study can be considered an initial step in understa nding the ways with which expertand community-governance can produce high quality knowl edge. Further, these propositions pave the way toward a theory of governance for elec tronic repositories, and provide a theoretical framework as a basis for future researc h.
69 A salient issue that deserves further discussion ab out these propositions concerns the effects posited in the propositions: the aspec ts of expertand community-governance contribute to knowledge quality only through main effects. This is because empirical data only provides evidence for main effects but no t for more complex relationships such as interaction effects. However, the paradigm mod el that is employed during axial coding classifies the aspects of expertand commun ity-governance as actions and conditions Consequently, the governance functions of both m echanisms are considered actions that increase knowledge quality, and the remaining aspects (i.e., experts credibility and experts ownership of contents for expert-governance; and communitys involvement for community-governance) are considere d conditions for achieving high quality knowledge in repositories. Therefore, the paradigm model employed during data analysis implies an interaction effect, where actions lead to outcomes contingent upon the necessary conditions This is intuitive because execution of the gover nance functions (i.e. actions ) alone may not necessarily translate into high qua lity knowledge (i.e., consequence ) without the credibility of experts or experts ow nership of contents (i.e., conditions ) in the case of expert-governance. Therefore, it is incumbent on future researchers to investigate the possibility of inter action effects among the aspects of a governance mechanism. To do so, studies should be designed incorporating the organizational level of analysis to capture both ac tions and conditions from a variety of organizational settings to examine the main and int eraction effects of the constructs proposed in this study. The second theoretical contribution of this study i s made to the literature on data and information quality. Quality is a rather nebul ous concept, and researchers have been
70 trying to understand the different dimensions of qu ality in a variety of contexts, including KM. The most commonly used framework in this domai n is the one developed by Wang and Strong (1996), which identifies four dimensions of quality for raw data These dimensions were later extended to processed data, o r information which signals the generalizability of the framework (c.f., Lee et al. 2002). This essay suggests that these dimensions are also applicable in the KM context. However, this finding should be interpreted cautiously as it does not conclusively show that dimensions of data and information quality are also applicable to knowledge quality. This study did not set out to make a clear-cut distinction between knowledge and information Therefore, this study contributes to the literature by suggesting that th e quality of articulated data (in the form of information or knowledge as opposed to raw data ) can be assessed using the existing dimensions of quality developed for raw data. Addi tionally, the quality of articulated data can be further conceptualized using a new dime nsion that concerns the application of the articulated data in a specific context. Thi s suggests that as researchers move higher in the data-information-knowledge hierarchy, additi onal new dimensions may be needed to articulate a more comprehensive representation o f the quality concept. Future research can further investigate this new dimension to exten d our current understanding of quality. This study, and interest in governance in general, is also expected to stimulate future research in KM. To the best of our knowledg e, this is one of the first studies that discusses different types of governance mechanisms as means to assess knowledge quality in organizational repositories. Although g overnance mechanisms are ubiquitous in many organizations, there is little appreciation of the concept of governance in KM. Many additional research questions besides the one examined in this study will be
71 stimulated. For instance, there is evidence in the sociology and organizational behavior literature that governance mechanisms can alter the way individuals behave in certain contexts (Adler and Borys, 1996; Bowles and Gintis, 2002; Streeck and Schmitter, 1985). As an example, hierarchical control and community-g overnance can cause negative attitudes and dissatisfaction in certain contexts, and thus result in withdrawal behaviors; whereas they can cause positive attitudes and thus citizenship behaviors in other contexts (Adler and Borys, 1996). Therefore, it is possible that expertand communitygovernance can induce individuals to behave differe ntly in organizational settings when providing contributions to or using knowledge from repositories. The paucity of studies in this area warrants the examination of knowledge contribution and knowledge use behaviors as elements of governance mechanisms (whi ch are investigated in the second and the third essays of this dissertation, respecti vely). Additionally, future research should investigate th e quality implications of the two governance mechanisms from an agency theory per spective. Since it is not possible to observe experts governance behaviors, expert-go vernance is susceptible to the agency problem. This is because it is difficult, if not i mpossible, for knowledge users to know whether experts execute governance functions, wheth er governance functions are executed thoroughly, and whether experts are credib le or have feelings of ownership toward repository contents. From this standpoint, it would be interesting to examine how knowledge users make judgments about these aspects of expert-governance, and how organizations can manipulate the related perception s of knowledge users. This is important, because if organizations can ensure that knowledge users have favorable perceptions, the use of knowledge from repositories can be further increased.
72 Consequently, future research might examine the way s with which expert-governance can be rendered more transparent to knowledge users For example, researchers could examine whether organizations should publicize the policies and procedures employed by experts, or report the metrics of governance proces ses to knowledge users. Researchers could also examine whether interactions between exp erts and knowledge contributors during the revision cycle increase the transparency of the governance processes, and whether these interactions create perceptions of ex perts credibility and experts ownership of content on the part of knowledge users Since community-governance is relatively transparen t from an agency perspective (as it provides all the governance related metrics such as edits, comments, revisions, changes, etc. publicly), future research could fo cus on the effectiveness of communitygovernance on improving knowledge quality. In doin g so, researchers might investigate the ways with which individuals motivation to exec ute governance functions and their involvement in governance processes can be increase d. Finally, future research could also test the propos itions offered in this study. This will require the development of a measurement instr ument with good psychometric properties. The instrument should measure the thor oughness of governance functions, credibility of experts, and ownership of contents f or expert-governance; and the continuous execution of governance functions, diver sity of members, and involvement of community members for community-governance. Some o f these constructs, such as credibility (e.g., Pornpitakpan, 2004), and involve ment of individuals (e.g., Zaichkowsky, 1985) have valid measurement items in the literatur e. Others, such as ownership,
73 thoroughness, and diversity of members will need re liable and valid items to underpin future work in this domain. Practical implications This study has several practical implications. Fir st, it informs practitioners by identifying the fundamental building blocks of two different governance mechanisms, namely expertand community-governance, that are u sed to improve knowledge quality in organizational repositories. Given the paucity of studies in this area, this will enable practitioners to make better decisions in implement ing a specific governance mechanism in their organizations. Specifically, the characte ristics of the two governance mechanisms discussed in this essay can be used to d etermining the mechanism that optimizes the use of KM in a specific organization. Second, this essay informs software development eff orts in organizations. Since governance mechanisms are instantiated partly by te chnological features, development teams should determine the type of governance mecha nism that will be used for the new repository during the requirements gathering phase to include those technological features associated with that specific mechanism. This is important since not paying attention to certain features might lead to the int roduction of forms of governance for which the organizational members do not have a good understanding. For example, if repositories are designed to enable knowledge users to provide feedback about existing contributions or to edit them, the repository might impose community-governance. However, if community-governance is not promoted ap propriately in the organization, or if the organizational culture is not ready to embra ce such a mechanism, employees might
74 reject the repository or fail to execute governance functions, both of which would hinder knowledge transfer efforts. Further, software development efforts should focus on increasing the transparency of expert-governance, as expert-governance can suff er from agency problems. Specifically, developers should incorporate meta-da ta about governance functions into the user interface to inform knowledge users about the extent of governance functions carried out on contributions. To further increase transparency, developers could publicize the governor (i.e., expert) of each contr ibution by first providing an identifier for each expert (such as first and last name), and then linking this identifier to the experts personal profile to inform knowledge users about the experts credibility and ownership of the content. The third and final practical implication of this s tudy is to enable practitioners to increase the efficacy of expertand community-gove rnance process and thus increase knowledge quality. In the case of expert-governanc e, organizations should ensure that (1) controls and checklists exist oblige experts to execute governance functions thoroughly, in the proper order, within a reasonabl e amount of time, and with appropriate diligence; (2) credible individuals, who have exten sive knowledge and experience in their domains, are designated as experts to execute the g overnance functions; and (3) feelings of ownership on the part of experts are engendered by repository contents through giving experts control over what to publish in repositorie s, and holding them responsible for the positive as well as the negative consequences of pu blished content, and (4) the precautions embodied in the previous three points a re communicated clearly to knowledge users to reduce agency problems. These f our measures may not only help
75 increase the quality of contributions stored in rep ositories, but also induce users to have more favorable perceptions toward these contributio ns, adding momentum to the quality improvement process. In the case of community-governance, organizations can increase efficacy of KM by ensuring that (1) governance functions are execu ted continuously and by a diverse set of members; and (2) community members have a high l evel of involvement in the governance process. The former can be achieved by pushing the contents of a repository periodically to employees through email or really simple syndication (RSS) to inform them of new or dated contributions in their domains. Employees might then be asked to look at these contributions, make necessar y changes, or provide reviews or comments. The latter might be achieved by encourag ing community members to execute governance functions on a regular basis. For insta nce, editing, reviewing, and rating activities could be incorporated into employees an nual performance measures, or they might be considered contributions made to reposit ories and rewarded using existing reward structures.
76 ESSAY II: USERS MOTIVATIONS TO CONTRIBUTE TO EXPER TAND COMMUNITY-GOVERNED REPOSITORIES Introduction Despite the prevalence of expertand community-gov ernance in many organizations, no study in the literature to the best of our knowledge differentiates between these two mechanisms in explaining the moti vations for making contributions to electronic repositories. The goal of this essay is to understand whether individuals motivations to contribute to expert-governed reposi tories differ from their motivations to contribute to community-governed repositories, and if yes how. Therefore, the specific research question of interest to this essay is: wha t factors influence individuals to make voluntary contributions to expertand community-go verned repositories? This essay is motivated by the fact that current li terature adopts a rather narrow perspective and explains motivations to make contri butions to repositories without taking governance mechanisms into account. Since reposito ries can be governed with different types of mechanisms (such as expertor community-g overnance), we need to refine our current understanding, and identify the factors tha t motivate individuals to contribute to expert-governed repositories compared to communitygoverned repositories. This is important, because governance literature suggests t hat different forms of governance induce different types of behaviors on the part of the governed. For instance, Adler and Borys (1996) argue that the degree of fit between the governance mechanism and the
77 context in which the governance mechanism is instan tiated determines whether individuals exhibit withdrawal or citizenship behav iors. For this reason, it is expected that different types of factors should motivate org anizational members to contribute to expertand community-governed repositories conting ent upon personal and contextual differences. However, the extant literature in KM does not provide much insight about the nature and the extent of these differences. Motivated by this gap in the literature, this essay conducts qualitative research using an interpretive paradigm to first identify th en compare the factors that motivate individuals to voluntarily contribute to expertan d community-governed repositories. The essay employs grounded theory to analyze the em pirical data collected from organizational members in a range of organizations. The research question is investigated for two different contexts, one in whi ch organizations use only one type of repository (either expertor community-governed), and another in which the expertand community-governed repositories are used simultaneo usly. This rest of this essay proceeds as follows. In th e next section, prior research in KM about contribution behaviors is reviewed. The f ollowing section presents the research methods used in this essay, which explains data collection procedure, sample characteristics, and data analysis. The next secti on presents the findings of this essay followed by the trustworthiness of findings. The f inal section summarizes key findings and discusses the theoretical and practical implica tions. Prior Research Explaining contribution behaviors has been a long-t ime goal for many researchers in the field of KM. As there exists a large body o f research in this area, current research
78 is synthesized using an input-process-output (IPO) framework (e.g., Hackman and Morris, 1975). In this framework, input represents the set of independent variables used to explain contribution behaviors, process represen ts the perspective used to explain how these variables influence contribution behaviors, a nd output represents the dependent variables used in the literature. There are many inputs (i.e., independent variables) investigated in the literature as potential determinants of contribution behaviors. Some of these variables are presented in Table 10 organized under five categories: (1) individual factors, which represent characteristics, beliefs, attitudes, and expectatio ns of individuals; (2) organizational factors, which represent characteristics of sponsoring orga nizations; (3) technological factors, which represent characteristics of the technologic al designs of knowledge repositories; (4) task related factors which represent the characteristics of organizati onal tasks performed; and (5) knowledge related factors which represent characteristics of knowledge. Among these factors, researchers focus mostly on individual factors as the primary determinant of contribution behaviors. Due to the breadth of individual factors examined in the literature, Table 10 includes only those individual factors that are examined by two or more studies. Concerning processes, prior literature uses three t ypes of perspectives, namely cognitive affective and social, to explain contribution behaviors. Cognitive proc esses explain contributions through contributors reasoni ng and rationality, and suggest that individuals make contributions because of certain e xpected outcomes (either for themselves or for the organization). Affective pro cesses are less rational in that they study contributors emotions, feelings, moods, and preferences to explain contribution
79 behaviors. Social processes, on the other hand, ex plain contribution behaviors through individuals interaction and socialization with eac h other, and suggest that social norms, influence, or obligations are drivers of contributi ons. Two examples studies illustrate the use of these th ree perspectives: Chiu et al. (2006) and Wasko and Faraj (2005). Using social ca pital theory as the underlying theoretical framework, both studies suggest that in dividuals make contributions because they expect to gain reputation in their organizatio n (a cognitive process); because they enjoy and feel good about helping others (an affect ive process); and because they feel obligated due to reciprocity and social norms (a so cial process). Table 11 summarizes the use of these processes in the literature along with the theoretical frameworks used by researchers. Three most commonly investigated outputs (i.e., dep endent variables) in the literature are: (1) intentions to make contribution s; (2) quality of contributions; and (3) quantity of contributions. The definitions and mea surements of these constructs are presented in Table 12. As seen in the table, inves tigations concerning quality of contributions are not as much as intentions or quan tity of contributions. Researchers focus mostly on quantity (i.e. volume) of contribut ions, which is measured through either self-reports or server-logs.
80 Definition Main effect (Study) Moderated by (support, study) Individuals factors Trust The belief in the good intent, competence, and reliability of employees/users with respect to contributing and using knowledge (Kankanhalli et al. 2005). Not supported (Chiu et al. 2006) Codification effort (supported, Kankanhalli et al., 2005) Identification The perception of similarity of values, membersh ip, and loyalty with the organization/community (Kankanhalli et al. 2005). Positive (Chiu et al., 2006) Positive (Dholakia et al., 2004) Positive (Bagozzi and Dholakia, 2002) Organizational reward (not supported, Kankanhalli et al., 2005) Reciprocity The belief that contributing to a repository will lead to a future request for knowledge being met (Kankanhalli et al. 2005). Positive (Chiu et al., 2006) Positive (Kankanhalli et al., 2005) Not supported (Wasko and Faraj, 2005) Social norms (supported, Kankanhalli et al., 2005) Need for reputation The need for receiving public apprecia tion and being recognized by others (Wasko and Faraj 2005). Positive (Kankanhalli et al., 2005) Positive (Wasko and Faraj, 2005) Social norms (not supported, Kankanhalli et al., 2005) Enjoyment in helping others (i.e., altruism) The pleasure obtained from helping others through contributing knowledge to a repository (Wasko and Faraj 2000). Positive (Kankanhalli et al., 2005) Positive (Wasko and Faraj, 2005) Personal outcome expectations Personal benefits that are expected to be obtained after making contributions to a repository (Chiu et al. 2006). Not supported (Chiu et al., 2006) Positive (Lin and Huang, 2008) Not supported (Yuan et al., 2005) Attitude toward knowledge sharing The degree of ones positive feelings about sharing knowledge (Bock et al. 2005). Positive (Bock et al., 2005) Positive (Chow and Chan, 2008) Not supported (Bagozzi and Dholakia, 2002) Positive (He and Wei, 2009) Social norm The degree to degree of perceived social pres sure to make contributions to a repository (Chow and Chan 2008). Positive (Bock et al., 2005) Positive (Chow and Chan, 2008) Not supported (Bagozzi and Dholakia, 2002) Self-efficacy The belief that individual himself/herself can prov ide valuable knowledge to the repository (Kankanhalli et al. 2005) Positive (Kankanhalli et al. 2005) Organizational commitment, organizational instrumentality, connective efficacy (supported, Kalman et al., 2002)
81 Technological comfort/competence The level of skills expertise in using electronic repositories (Yuan et al. 2005). Positive (Jarvenpaa and Staples, 2000) Positive (Yuan et al., 2005) Organizational factors Organizational reward Incentives provided for knowledge con tributions (Kankanhalli et al. 2005). Positive (Kankanhalli et al., 2005) Information culture Values and attitudes toward information, in formation processing, publishing, and communication (Jarvenpaa and Staples 2000). Positive (Jarvenpaa and Staples, 2000) Organizational ownership of information The degree to which individuals perceive as information belongs to organization rather than themselves (Jarvenpaa and Staples 2000). Negative (Jarvenpaa and Staples, 2000) Positive (Constant et al., 1994) Organizational climate The perception that organizational practices are fair and equitable (Bock et al. 2005) Positive (Bock et al., 2005) Technological factors IT infrastructure quality Degree to which the infrastructure of the repository meets members expectations with respect to response time, user-interface, etc. (Koh et al. 2007) Leaders involvement, level of offline interaction, usefulness (not supported, Koh et al., 2007) Task related factors Task interdependence The degree to which organizational task s depend on each other (Lin and Huang 2008) Positive (Jarvenpaa and Staples, 2000) Positive (Lin and Huang, 2008) Knowledge related factors Knowledge characteristics The perceived quality, accessibility, cost, and use of knowledge (Jarvenpaa and Staples 2000) Positive (Jarvenpaa and Staples, 2000) Table 10. A sample of independent variables investi gated in the literature
82 Study Cognitive Process Affective Process Social Process Theory Used Chiu et al. (2006) X X X Social capital theory Wasko and Faraj (2005) X X X Lin and Huang (2008) X Task-technology fit Cosley et al. (2005) X Collective effort model Yuan et al. (2005) X Collective action Cummings et al. (2002) X Social exchange theory Jarvenpaaa and Staples (2000) X X X Kankanhalli et al. (2005) X X X Koh et al. (2007) X Constant et al. (1994) X Bock et al. (2005) X X X Theory of planned behavior Chow and Chan (2008) X X X Bagozzi and Dholakia (2002) X X Dholakia et al. (2004) X X X Kalman et al. (2002) X X X Expectancy theory Chen (2007) X X X Expectation-confirmation theory He and Wei (2009) X Table 11. Processes identified in the literature Besides the IPO framework, it is also important to examine whether prior research differentiates between governance mechanisms in inv estigating contribution behaviors. The cross-tabulation in Table 13 shows that other t han a few exceptions the majority of studies do not report the type of governance mechan ism used in repositories. This indicates that that prior research does not take go vernance mechanisms into consideration when explaining contribution behaviors. Of the thr ee studies that mention the type of governance mechanism, Kalman et al. (2002) and Cumm ings et al. (2002) investigate self-governed repositories, while Cosley et al. (20 05) study participation behaviors in a non-organizational community-governed repository. It is noteworthy that studies that examine general knowledge sharing behaviors rather than contributing to electronic
83 repositories are not included in the table (e.g., B ock et al., 2005; Chow and Chan, 2008; Constant et al., 1994). Dependent variable Definition Measurement Study Intentions to make contributions Individuals willingness to make contributions to a repository Self-reported Bagozzi and Dholakia (2002) Bock et al. (2005) Constant et al. (1994) Chen (2007) Chow and Chan (2008) Kalman et al. (2002) He and Wei (2009) Quality of contributions Helpfulness of contributions (i.e., providing a direct answer and its source) Content analysis Wasko and Faraj (2005) Relevance, ease of understanding, accuracy, completeness, reliability, and timeliness of contributions Self-reported Chiu et al. (2006) Correctness of contributions Simple count of correct entries Cosley et al (2005) Quantity of contributions Volume of contributions Either self-reported or based on server-logs Chiu et al. (2006) Cosley et al. (2005) Cummings et al. (2002) Dholakia et al. (2004) Jarvenpaa and Staples (2000) Kankanhalli et al. (2005) Koh et al. (2007) Lin and Huang (2008) Wasko and Faraj (2005) Yuan et al. (2005) Table 12. Dependent variables investigated in the l iterature Prior literature provides two key insights: (1) no single theory may adequately explain contribution behaviors, but several differe nt perspectives may be integrated to achieve sufficient levels of explanatory power; (2) contribution behaviors are not solely determined by individual factors, but by organizati onal, technological, task related, and knowledge related factors as well.
84 Type of repository Governance mechanism Organizational Non-organizational Expert-governance ( ) ( ) Community-governance ( ) Cosley et al. (2005) Self-governance Kalman et al. (2002) Cummings et al. (2002) Not mentioned Jarvenpaaa and Staples (2000) Yuan et al. (2005) Kankanhalli et al. (2005) Lin and Huang (2008) He and Wei (2009) Bagozzi and Dholakia (2002) Dholakia et al. (2004) Wasko and Faraj (2005) Chiu et al. (2006) Koh et al. (2007) Chen (2007) Table 13. Types of repositories studied and their g overnance mechanisms Despite these insights, prior literature does not t ake governance mechanisms into account in explaining contribution behaviors. Alth ough mechanisms such as expertgovernance and community-governance are commonly us ed in organizations, there are no studies that distinguish between individuals mo tivations to make contributions to repositories governed by these two types of mechani sms. This essay attempts to address this gap in the literature, and adopts an interpret ive paradigm to building models of contribution behaviors using qualitative research. Research Methods This essay uses the same research methodology outli ned in the first essay. It conducts qualitative research using the interpretiv e paradigm, and uses grounded theory to address the research question of interest. The motivation to choose this research perspective is similar to the first essay in that p rior literature does not provide a priori
85 theories to undertake quantitative research using a positivist paradigm. In order to eliminate redundancy, the research methods employed for this study is not repeated. Readers can refer to the Research Methods section o f the first essay to find more information about the research methodology. Data collection procedure The data for the first and second essays were colle cted at the same time. Therefore, the same data collection procedure outli ned in the first essay was used to address the research questions of the second essay. For this reason, readers are advised to refer to the Data Collection Procedure section of t he first essay for more information about how data were collected. One difference between the data collection procedur es of the two essays was that, after identifying the governance mechanism(s) emplo yed in each participants organization, the second essay used the critical in cident technique (Flanagan, 1954) to elicit responses specific for that governance mecha nism. According to this technique, two different incidents were defined: (1) making a contribution, and (2) not making a contribution to the repository employed in the part icipants organization. Therefore, the data collection instrument asked each participant t o recall the last substantial contribution he/she made (and could have made but did not make) to the repository being used in his/her organization, and briefly describe the natu re of this contribution. Following the description of the incident, each participant was a sked probing questions about his/her motivation for making the contribution in the first incident, why he/she did not make the contribution in the second incident, and if appli cable whether he/she could have made
86 the same contribution in the first incident to the other repository that used the alternative governance mechanism, and why or why not. Although each question had a comment box for partic ipants to type their answers, several of the probing questions also included precoded items to choose from. These items were identified from prior studies in the lit erature, and were included in the questionnaire to reduce the typing cost of particip ants and minimize their fatigue. For example, when participants were asked about their m otivation for making their last contribution, there were four pre-coded items to ch oose from, which included: (1) to gain reputation in my organization, (2) for altruism, (3 ) for reciprocity, and (4) for organizational rewards. The screenshot presented i n Figure 4 further shows the design of this particular question. Figure 4. Screenshot of an example question
87 As described in the Data Collection Procedure secti on of the first essay, a second phase of the data collection was undertaken online. The questions in both the first and the second phases were the same, except the questio ns in the second phase included insights gained from the first phase. Specifically some of the answers identified as being salient in the first phase were pre-coded as possib le answers in the second phase. This was motivated by two reasons. First, as suggested by Flanagan (1954), there was a need to determine whether the responses collected in the first phase were general behaviors or were highly specific to the described incidents. S econd, pre-coded answers reduced the fatigue, and thus, the drop-out rate of participant s. This was necessary, because fatigue and the time required to complete the questionnaire acted against getting usable answers from participants. As an example to demonstrate how these pre-coded it ems were developed and included into the second phase, consider the questi on discussed earlier in Figure 4 about the motivations of participants to make contributio ns to repositories. The analysis of the responses collected in the first phase suggested th at there were three additional reasons why participants made contributions: (1) for reason s that would benefit my organization, (2) for reasons that would benefit myself, and (3) to fulfill my job responsibilities. When the same question was asked to participants in the second phase, these three reasons were added to the existing pre-coded items as presented in Figure 5. The findings section discusses the use of these pre-coded items and part icipants corresponding responses whenever applicable.
88 Figure 5. Screenshot of an example interview questi on in the second phase Sample characteristics Since the first and the second essays used the same sample, readers can refer to the Sample Characteristics section of the first ess ay for more information about the demographics and characteristics of participants. Data analysis The data collected from participants were examined using open, axial, and selective coding (Strauss and Corbin, 1998). In or der to demonstrate the data analysis procedure, the coding process for one of the factor s that motivated participants to make contributions to expert-governed repositories is ex plained below. As the first step, comments related to reasons for providing contribut ions to expert-governed repositories
89 were identified in the data set. These comments we re usually short statements, such as those presented in Table 14, that were typed into c omment boxes provided for the related question in the online questionnaire. In the second step of data analysis, open coding wa s performed using a line-byline analysis to identify concepts in the comment s. For example, as seen in the first comment in Table 14, two concepts were identified u sing in vivo codes as highlighted in the original response: reducing time and increasing team effectiveness Similarly, in the second comment, the participant mentioned that he/s he provided a contribution to the expert-governed repository to improve quality and customer experience (as highlighted in the original text). Participant comment Concepts Category To reduce [the] time to solve a problem and thereby increase [the] overall effectiveness of our team. Reducing time, increasing team effectiveness Organizational benefits To improve quality and provide best customer experi ence Service quality, customer experience To standardize budget processes for [next year] Process standardization To improve the quality of my team's services to cli ents and [to other] areas of the [firm]. High quality service (internal & external customers) The current economic environment has forced me to r eally analyze my business and marketing strategies. New strategy Table 14. Participant comments for providing contri butions to expert-governance After concepts were identified for each comment, th e similarities and differences between these concepts were examined to create high er order categories (hereafter referred to as factors). For example, the simila rity between the concepts identified in Table 14 was that they were all organizational outc omes. In other words, the participants
90 were providing contributions to expert-governed rep ositories with the expectation that these contributions would benefit certain aspects o f their organizations. Therefore, these concepts were grouped under the organizational benefits factor. The above example demonstrates how one factor was i dentified using open coding for making contributions to expert-governed repositories. Applying the same technique to the rest of the data generated many mo re concepts, and thus factors, as presented in Table 15. Expert-governance Concepts Factors Gaining personal benefits, enhancing work life, ease of locating information Personal benefits Volunteer, helping, personal satisfaction Altruism Familiarity with the process, not enough time Codification effort Limited knowledge, new to position Lack of expertise Similar contributions Risk of duplication Table 15. Concepts and categories identified for en ablers of expert-governance Following open coding, axial coding was performed t o identify relationships between factors, and understand how the factors fit the paradigm model proposed by Strauss and Corbin (1998). Strauss and Corbin sugg est that during axial coding researchers should define a paradigm model that con sists of actions conditions and consequences in the study context, and try to identify which fa ctors map onto this model. The reason for using this model is that actions and conditions make up the ingredients for
91 consequences, and thereby, help researchers develop hypotheses about the observed phenomenon. In the context of this study, the consequence aspect of the paradigm model was set a priori by the questions during data collection. For exam ple, the comments used for organizational benefits (for which the coding proce dure was demonstrated earlier) were provided in response to the question what was your motivation for making that contribution , which set the consequence aspect of the paradigm model to making contribution Although not explicitly stated in the question, the fact that this question was asked for expert-governed repositories develope d a priori relationships between any of the factors identified from the responses and making contributions to expert-governed repositories According to the Strauss and Corbins paradigm mode l, there has to be an action that triggers the consequence. In the context of t his study, this action can at best be the act of codification before making a contribution. Therefore, if conve rting tacit knowledge to explicit knowledge represents the act of codification (i.e., the action), then sharing the codified knowledge through an electroni c repository (which is the dependent variable of interest to this study) represents making contributions (i.e., the consequence). This conceptualization suggests that the factors id entified in open coding represent the necessary conditions that facilitate the action. Therefore, the hierar chical relationship, presented in Figure 6, depicts the paradigm model, where organizational benefits is axially related to making contributions to expert-g overned repositories.
Figure In the last step of the coding process, selective c oding was used to put together all the relationships identified in axial coding. Acco rding to Strauss and Corbin se lective coding is the process of developing a unify ing story around a central factor to address the research questions. were set a priori by the questions in the data collection instruments These involved making contributions to two contexts: when there is no alternative reposito ry, and when there is an alternative repository with the other governance mechanism. factor was making contributions to expert alternative repositories) In order to show the relationships between facto rs identified in this study and the central factors, diagramming coding, as proposed by Strauss and Corbin Note that, the procedure identified as a salient driver of making contributi ons to expert Applying the same procedure to the rest of the data helped identify many more factors for 92 Figure 6. Hierarchical structure of categories In the last step of the coding process, selective c oding was used to put together all the relationships identified in axial coding. Acco rding to Strauss and Corbin lective coding is the process of developing a unify ing story around a central factor to research questions. This essay used several different central factor by the questions in the data collection instruments These making contributions to a repository with a specific governance mechanism two contexts: when there is no alternative reposito ry, and when there is an alternative repository with the other governance mechanism. For the above exa mple, the central factor was making contributions to expert -governed repositories ( when there were no In order to show the relationships between facto rs identified in this study and the central factors, diagramming technique was used during selective proposed by Strauss and Corbin (1998). procedure explained above demonstrates how one factor was identified as a salient driver of making contributi ons to expert governed repositories. Applying the same procedure to the rest of the data helped identify many more factors for Making contributions to expert-governed repositories Codification Organizational outcome expectations(Action) (Condition) (Consequence) In the last step of the coding process, selective c oding was used to put together all the relationships identified in axial coding. Acco rding to Strauss and Corbin (1998), lective coding is the process of developing a unify ing story around a central factor to central factor s, which by the questions in the data collection instruments These central factors governance mechanism in two contexts: when there is no alternative reposito ry, and when there is an alternative mple, the central when there were no In order to show the relationships between facto rs identified in used during selective demonstrates how one factor was governed repositories. Applying the same procedure to the rest of the data helped identify many more factors for
93 both expertand community-governed repositories. The next section presents the findings and provides the related evidence. Findings Since the research question of this study is invest igated in two different contexts, the findings of this study are presented separately for these contexts in the below two sub-sections. Existence of one governance mechanism In order to understand the factors that motivated i ndividuals to contribute to expertand community-governed repositories, partic ipants who used only one type of repository were identified in the data set. The re sponses of these participants for two questions (i.e., what was your motivation for that contribution and why did you not make that contribution ) were analyzed separately for both expert-governe d and community-governed repositories. The reason for ad opting such a methodology was to compare the factors that were identified for expert -governed repositories with those identified for community-governed repositories. Th e analysis of the data suggested that the factors that explained contribution behaviors f or expert-governed repositories were the same as those for community-governed repositori es with two exceptions. A side-byside comparison between the two types of repositori es is presented in Figure 7. As seen in the figure, organizational benefits, rep utation, altruism, and organizational rewards were positively related to m aking contributions for both expertand community-governed repositories. However, pers onal benefits, as one of the factors for expert-governed repositories, was not observed for community-governed repositories; and reciprocity, as one of the factors for communit y-governed repositories, was not
94 observed for expert-governed repositories. Further more, codification effort, lack of expertise, and risk of duplication were identified as factors that were negatively related to making contributions to both expertand communitygoverned repositories. Factors that explain making contributions to expert-governed repositories Factors that explain making contributions to community-governed repositories Figure 7. Comparison of factors identified for expe rtand community-governed repositories The differences between the two models, as seen in Figure 7, are interesting. Although not observing the effects of personal bene fits and reciprocity for the alternative repositories can be a sample-specific finding, the findings can also indicate the emergence of a new conceptualization for explaining contribution behaviors. Personal benefits is more related to self-development, where contributions help individuals
95 improve their future performance. On the other han d, reciprocity is a manifestation of social exchange, where individuals make contributio ns to fulfill their obligations from an earlier help they received or to get help from othe rs in the future. This indicates that expert-governed repositories serve individuals sel f-development needs, whereas community-governed repositories promote social exch ange. The descriptions of each of these factors are presented below with the correspo nding evidence for their existence. Organizational benefits In the context of this study, organizational benefi ts can be defined as organizational gains from providing contributions t o repositories. Contributions can provide many benefits to organizations, some of whi ch include increased efficiency, effectiveness, or capacity. Contrary to the notion that individuals seek their self interests, previous literature has reported that employees mak e contributions to repositories in the interests of their organizations as well (e.g., Chi u et al., 2006; Lin and Huang, 2008). The data collected in the first phase of the study provided support for this argument for both expertand community-governed repositories, s ince interviewees mentioned that their motivations to contribute were, To reduce [the] time to solve a problem and thereb y increase overall effectiveness of our team To improve quality and provide best customer exper ience To standardize budget processes for [next year] To improve the quality of my team's services to cl ients and [to other] areas of the [firm] Similar comments were made for community-governed r epositories as well. One interviewee suggested that her motivation to contri bute stemmed from the need to
96 standardize the current process and create document ation for future use (which is expected to reduce future inefficiencies in executi ng the same process). She commented that her motivation was, to ensure there is documentation for what the rule s [are] and what is being implemented One interviewees commented that her contribution to the community-governed repository in her organization was intended to increase the capacity of one of the frontoffice processes. The salience of organizational benefits became more evident in the second phase of the study. When individuals were offered the ch oice to select organizational benefits as their motivation to contribute (which was coded as for reasons that would benefit my organizations, [e.g., make us more productive]), 3 8% of participants (or 11 out of 29) who made contributions to expert-governed repositor ies, and 34% of participants (or 10 out of 29) who made contributions to community-gove rned repositories selected organizational benefits as an the reason for their latest contribution. Based on the above discussion, this study proposes, P1: Organizational benefits are positively related to providing contributions to both expertand community-governe d repositories. Reputation Reputation is defined as individuals perceptions o f their self-image in the eyes of others (Kankanhalli et al., 2005). Being a reputab le individual at workplace has many benefits for employees. For example, individuals g ain respect from others, are treated as experts who know everything, and have a better ch ance of getting promoted or securing
97 their jobs in their organizations (Constant et al., 1994; Wasko and Faraj, 2000). While providing contributions to repositories is one of t he methods for building reputation in organizations, it can be considered an effective me thod since contributions reflect the extent of expertise possessed by individuals. Drawing upon the previous findings in the literatur e, reputation was provided as a pre-coded response to participants in both the firs t and the second phases of data collection. The results showed that 22% (or 11 out of 50) interviewees who used expertgoverned repositories, and 21% (or 8 out of 39) par ticipants who used communitygoverned repositories chose the option of gaining reputation as one of the drivers of their latest contribution. It is also interesting to note that the interview data provided evidence for the relationship between reputation an d contribution behavior in the negative direction, where individuals desire to be less reputable (and thus less visible) in an organization may lead to abstaining from making contributions to repositories. For example, one interviewee, who tried to explain why she did not provide a contribution to the expert-governed repository in her organization, commented, I did not want to get selected as an expert in [th is area] because that work environment can be high-stress. While [I ] may be interested in working in this area in the future, I was not interested in taking on that type of client while doing the MB A program Therefore, individuals need or desire to build rep utation in an organization acts as a salient driver of making contributions to repo sitories regardless of the type of governance mechanism used in those repositories. T his leads to proposing, P2: Gaining reputation is positively related to pro viding contributions to both expertand community-governe d repositories.
98 Altruism Altruism is defined as individuals desire to help others. Altruistic motivations for providing contributions to repositories has bee n conceptualized in prior literature as enjoyment to help others (Kankanhalli et al., 2005; Wasko and Faraj, 2005). The fundamental premise of this construct is that indiv iduals provide contributions to repositories not because of any outcome expectation s or rewards, but out of goodwill and the sheer enjoyment of helping others. Based on the previous findings in the literature, a ltruism was provided as a precoded response to participants in both the first an d the second phases of data collection. The results showed that 24% of participants (or 12 out of 50) who used expert-governed repositories, and 26% of participants (or 10 out of 39) who used community-governed repositories chose the option of altruism as one of the drivers of their latest contribution. Besides the quantitative data, parti cipants also provided qualitative data about the motivational effect of altruism. For exa mple, one interviewee who contributed to an expert-governed repository commented that his /her motivation was to, Provide a mechanism for others to access informati on Another interviewee who contributed to an expert-go verned repository mentioned, This is what I am good at and I love to do it The data provided evidence for the motivational eff ect of altruism for communitygoverned repositories as well. For example, one pa rticipant commented, This effort was the idea of several people and I v olunteered to help perform the research Another mentioned that his/her motivation was to,
99 Help our customers and team members, and [it is] satisfying to help someone In line with existing research, the above findings suggest that altruism is a salient motivator for providing contributions to repositori es, and its salience does not depend on the type of governance mechanism used for repositor ies. This leads to proposing, P3: Altruism is positively related to providing con tributions to both expertand community-governed repositories. Organizational rewards Existing literature defines organizational rewards as incentives offered by organizations for providing contributions to knowle dge repositories (Ba et al., 2001; Kankanhalli et al., 2005). Rewards take different shapes and forms in different organizations, most common of which are bonuses, pa y increases, or promotions. In certain organizations contributions directly influe nce rewards, whereas in others they affect rewards indirectly through performance evalu ations. Based on the previous findings in the literature, organizational rewards was provided as a pre-coded item in the form of for organizational rewards for participan ts to choose. The data collected in both phases of this study rev ealed that organizational rewards was a salient factor for making contributio ns to both expertand communitygoverned repositories. For example, 16% of partici pants (or 8 out of 50) who made contributions to expert-governed repositories chose organizational rewards as the underlying reason for their latest contribution. O n the other hand, 15% of participants (or 5 out of 39) who used community-governed repositori es chose organizational rewards as the driver of their contribution behaviors. This l eads to proposing,
100 P4: Organizational rewards are positively related p roviding contributions to both expertand community-governe d repositories. Personal benefits For the purposes of this study, personal benefits i s defined as personal gains from providing contributions to repositories. It emphas izes that individuals make contributions to repositories to benefit themselves rather than b enefiting the organization, department, or unit. The concept of personal benefits is roote d in the social cognitive theory (Bandura, 1986), which suggests that individuals se ek their self-interests and are more likely to perform actions that benefit themselves. In the context of this study, the immediate benefits of making contributions are impr ovements in personal efficiency and effectiveness, enhancements in personal and profess ional development, and organization of personal knowledge. The data collected for this study suggested that pe rsonal benefits were important for making contributions only for expert-governed r epositories. When asked about the reason for their latest contribution, one participa nt who contributed the requirements of a systems analysis design project to his expert-gover ned repository, commented, Contributions directly enhance my quality of life and ease of acquiring information. I need to make sure that I have everything I need [to perform my task] Further evidence for the salience of personal benef its was observed in two rather general comments. In order to emphasize the import ance of providing contributions to repositories, two participants, who used expert-gov erned repositories in their firms, mentioned that contributions helped contributors re call certain intricacies of
101 organizational tasks, and increase personal efficie ncy. One of these participants, working in finance, mentioned, Our repository serves as the knowledge base which comes to [ones] own rescue many times, because the provider of the content would have forgotten the details of the con tribution after sometime. The other participant commented, [Contributions] help me perform better in my tasks [My] overall effectiveness and efficiency increases. Further support for personal benefits was provided by the second phase of the study. When participants were asked the reason for their latest contribution to the repository used in their firm, 34% of them (or 10 o ut of 29), who used expert-governed repositories, selected the pre-coded response that read as for reasons that would benefit myself (e.g., self learning, productivity, etc.). The support for the salience of personal benefits for community-governed repositories was ra ther weak. No one in the first phase provided any comments about personal benefits, and only two participants (out of 29) in the second phase chose the related pre-coded respon se as their motivation for contributing to community-governed repositories. While this may be a sample-specific finding, it may also be because of the characteristics of expertand community-governed r epositories. Expert-governed repositories provide a good place to organize perso nal knowledge that ultimately contribute to personal productivity and efficiency, because these types of repositories prevent others to tamper with contributions (throug h editing), or provide unsolicited feedback (through comments and ratings). For this reason, it is possible for expertgoverned repositories to attract more contributions as a result of individuals selfdevelopment efforts. This leads to proposing,
102 P5: Personal benefits are positively related to pro viding contributions to expert-governed repositories. Reciprocity Reciprocity is defined as the sense of mutual inde btedness (Wasko and Faraj, 2005, p.43). It induces individuals to maintain a sense of fairness in their relationships with others, and makes them provide contributions t o repositories either to fulfill their obligations from an earlier help they received, or to receive help from others when they need it in the future. The motivational effect of reciprocity on providing contributions has already been reported in the existing literature (Kankanhalli et al., 2005; Wasko and Faraj, 2005). In light of prior research, reciprocity was provided t o participants as a pre-coded response in both the first and the second phases of data collec tion. However, the data revealed that reciprocity was more salient for community-governed repositories than expert-governed repositories. For both phases of data collection, 26% of participants (or 10 out of 39) who contributed to community-governed repositories mentioned reciprocity as the underlying reason for their contribution, whereas o nly two participants (out of 50), among contributors of expert-governed repositories mentioned it as their motivation. Although the salience of reciprocity for only comm unity-governed repositories can be an artifact of the small sample size used in this study, it may also be an implication of community-governance. As discussed earlier, community-governance enables more interaction among organizational membe rs. When individuals interact, reciprocity overrides self-interest especially if i ndividuals know each other and are interdependent to one another (Axelrod, 1984) such as in communities. Therefore,
103 making contributions to repositories is governed th rough the norm of social exchange rather than the notion of self-interest (as in expe rt-governed repositories). Individuals make contributions to either get help from others i n the future, or to fulfill their obligations from an earlier help they received. On the other hand, the effect of reciprocity is observed less in expert-governed rep ositories, as these repositories are less social. This further corroborates the argument tha t self-development (as in personal benefits) is more salient for expert-governed repos itories, and social exchange (as in reciprocity) is more salient for community-governed repositories. This leads to proposing, P6: Reciprocity is positively related to providing contributions to community-governed repositories. Codification effort Codification effort can be defined as the time and effort needed to make a contribution to a knowledge repository (Kankanhalli et al., 2005). Prior studies in the literature have reported a negative relationship be tween codification effort and contribution behaviors (Kankanhalli et al., 2005; M arkus, 2001). The analysis of the data echoed the same finding as codification effort nega tively influenced providing contributions to both expertand community-governe d repositories. For example, one interviewee could have contributed a process flow t o the expert-governed repository used in her organization, but the effort required for de scribing the steps, eliminating any ambiguities in the description, and formatting the document dissuaded her to do so. She commented that she could use the time to perform he r daily tasks and avoid staying late for overtime. In another instance, one interviewee stated that she did not make a
104 contribution to her organizations expert-governed repository, because her (and other stakeholders) familiarity with the process did not justify spending time and effort on codifying the intricacies of the process. The evid ence for the negative effect of codification effort for contributing to community-g overned repositories came from one participant, who mentioned that he/she failed to do cument a modification to one of the value-adding processes in his organization, because he/she was busy with other work and didnt have enough time for codification. This leads to proposing, P7: Codification effort is negatively related to pr oviding contributions to both expertand community-governe d repositories. Lack of expertise Expertise represents the extent of skills in a spec ific domain. It is independent of total work experience, and concerns the skills poss essed in a context over a period of time. For instance, an individual may have experti se on IT security, but the same individual may lack expertise in accounting and can quickly become a novice if asked to perform bookkeeping. Prior literature conceptualiz es expertise to have a positive effect on contribution behaviors, although it fails to sup port this relationship (e.g., Wasko and Faraj, 2005). An explanation for this inconsistenc y is that expertise does not necessarily induce individuals to provide more contributions, a lthough its absence certainly prevents contribution behaviors. The data provided support for this argument as participants mentioned that their lack of expertise prevented th em from making contributions to repositories regardless of the type of governance m echanism used for those repositories. For example, one interviewee indicated that he did not make any contributions to the expert-governed repository used in his firm, becaus e,
105 my level of knowledge in [my current area] is very limited In another instance, another interviewee who used a community-governed repository in his/her firm commented that she could not make any contribution, because she was new to position . Therefore, this study proposes, P8: Lack of expertise is negatively related to prov iding contributions to both expertand community-governe d repositories. Risk of duplication Duplication concerns the possibility of providing a contribution that is similar to existing contributions in knowledge repositories. When participants were asked to state their reasons for failing to provide contributions to repositories used in their organizations, they mentioned risk of duplication a s one of the underlying reasons. For example, one participant who used an expert-governe d repository in his firm mentioned, A similar contribution was already done by someone else. Among participants who used community-governed repo sitories, one commented, there is too much similar information in the commu nity-governed repository. The negative effect of duplication was more pervasi ve for community-governed repositories especially in the second phase of data collection. For example, three (out of 29) participants who used community-governed reposi tories mentioned the risk of duplication as the reason for not providing contrib utions, whereas no participant mentioned risk of duplication as the reason for con tributing to expert-governed repositories. However, the risk of duplication was observed for both expertand community-governed repositories to propose,
106 P9: Risk of duplication is negatively related to pr oviding contributions to both expertand community-governe d repositories. Although risk of duplication is a rather intuitive factor that negatively affects contribution behaviors, its prevalence in community -governed repositories is noteworthy. This finding may mean that community-governance is less likely to eliminate duplication (or organize similar types of information) compared to expert-governed repositories. This could be because of community members lack of involvement in executing the governance functions. Existence of two governance mechanisms As mentioned earlier, this essay examines the resea rch question in two different contexts: one in which there exists only one type o f repository (either expertor community-governed), and another in which the two t ypes of repositories exist simultaneously (both expertand community-governed ). This sub-section investigates the research question in the second context. The g oal is to understand the factors that induce individuals to make a choice between expertand community-governed repositories in making contributions. For this rea son, the responses of participants, who used the two types of repositories simultaneously i n their firms, were analyzed. These participants were asked the recall the last substan tial contribution they made to one repository (such as expert-governed), and discuss w hy they did not make the same contribution to the alternative repository (in this case community-governed), and whether the alternative repository would have been a better choice for making that contribution. The findings suggested that when expertand commun ity-governed repositories were used simultaneously, contribution behaviors we re influenced by two sets of factors:
107 (1) knowledge-based, and (2) need-based. Knowledge -based factors involved the type of contribution (i.e., whether the contribution was a suggestion/idea), and the characteristics of contribution (i.e., the degree of the formality and the sensitivity of the contribution). On the other hand, need-based factors involved part icipants need for collaboration, expert-validation, and recognition. The effects of these factors on making contributions to expertand community-governed repositories are depicted in Figure 8. Figure 8. Choice of governance mechanisms As seen in the figure, individuals were more likely to provide contributions to expert-governed repositories (compared to community -governed repositories), if those contributions were formal or sensitive or if individuals were in need of expert validation or recognition On the other hand, they were more likely to make contributions to community-governed repositories (compared to expert -governed repositories), if those contributions were suggestions / ideas or considered informal or individuals were in need of collaboration The concepts identified during open coding to id entify the above
108 factors are presented in Table 16. The rest of thi s section explains these factors and presents the evidence for their existence. Suggestions/ideas Suggestions and ideas represent recommendations tha t challenge the current (or introduce new) ways of doing business in organizati ons. For example, project proposals, suggestions for process flows, recommendations on h ow to solve existing problems, or new approaches in achieving the targeted outcomes c an all be considered suggestions/ideas contributed to knowledge reposito ries. Some of the contributions identified as suggestions/ideas in the data set inc lude a new project for a front-office business process, recommendations about software de velopment and implementation processes, and suggestions on revamping the sales e fforts. Concepts Factors Idea input, new idea, proposal, rough ideas or concepts Suggestions/ideas Formal approved communication, formal structure Informal contributions Reviewing, co-authoring, collaboration, alter Co-authoring and feedback Risky knowledge, regulated knowledge Sensitivity of knowledge Formal approved communication, formal structure Formality of contributions Expert vetting, polishing, experts increasing quality Expert validation Gain recognition Need for recognition Table 16. Concepts and categories identified for ch oice of governance mechanism
109 Participants who used expertand community-governe d repositories in parallel mentioned that were less likely to contribute sugge stions/ides to expert-governed repositories. For example, when asked the reason f or not providing the contribution to the expert-governed repository, one participant men tioned, [expert-governed repository] is not an area for id ea input. Another participant, who was a designated expert fo r the expert-governed repository used in her organization, commented, I havent received a new idea or proposal yet. I g uess people didnt submit anything like that yet. This view received support in the second phase of d ata collection as well. When individuals were asked why they did not make their latest contribution to an expertgoverned repository, 27% of them (or 6 out of 22) c hose the option which stated that expert-governed repository is not for contributing suggestions/ideas. Follow-up interviews revealed two major reasons for this. First, contributors believed that experts might not evaluate or even ap preciate the quality or the usefulness of suggestions/ideas contributed to expert-governed repositories. Expert-governance is very good at checking the accuracy of contributions validating them, and ensuring that they do not mislead knowledge users. However, when it comes to evaluating the value propositions of a new suggestion or an idea, expert -governance may not be the best option, since the processes used to vet contributio ns (such as performing a fact-checking, or putting the contributions to a test) may not app ly to these types of contributions. Therefore, it is possible for experts to undervalue suggestions/ideas and reject them, or overvalue them although they are not applicable in the field. One interviewee in the manufacturing industry provided support for this ar gument. Following the submission of
110 a new suggestion about the design of a specific par t to the expert-governed repository of the firm, experts published the suggestion only to find that none of the workers used it or even perceived it as valid. The second reason for individuals unwillingness to contribute suggestions/ideas to expert-governed repositories is that suggestions /ideas are usually considered work-inprogress products rather than finalized products. They may consist of concepts that have not been tested or validated in the field, whi ch create concerns for their validity and applicability. Therefore, even if they are submitt ed to expert-governed repositories, their likelihood of being rejected is very high. Just li ke academic manuscripts, they need to go through a ripening process, where they are founded on strong principles, are proven to work in the field, and are vetted by sufficient num ber of colleagues for their applicability. Therefore, it is not likely for individuals to subm it suggestions/ideas to expert-governed repositories, unless they ensure that these suggest ions/ideas can withstand the meticulous governance process imposed by expert-governance. F urther, expert-governed repositories may not be a good choice for the ripen ing period of suggestions/ideas, since the design of these repositories provides limited s upport for organizational members to collaborate with each other or provide feedback. C ollaboration and feedback are essential, as they help individuals incorporate dif ferent perspectives into the suggestions/ideas and improve their value and appli cability in the field. The above perspective was supported by one participant, who w as responsible for overseeing the expertand community governed repositories in the organization. He commented, People use the wiki much more when they are creati ng a new idea, or a point of view, or maybe an idea from a s ervice offering as an example. [Community-governed repository] is a place where people can collaborate around that and take v ery rough
111 ideas or concepts and sort of percolate them into s omething more tangible and formal. When they get a work product that they consider reusable, then they submit it to the [expe rt-governed repository]. Therefore, this study proposes, P10: When expertand community-governed repositor ies are used simultaneously, suggestions/ideas are (a) more like ly to be contributed to community-governed repositories, and (b) less likely to be contributed to expert-governed repositories. Sensitivity of knowledge In the context of this study, sensitivity of knowle dge represents the degree to which knowledge may have legal ramifications if cod ified (or used) inappropriately in an organizational setting. It connotes the risk associated with the inaccurate codification or inappropriate use of knowledge, both of which may c ause tangible or intangible damage to employees or organizations. For example, regula tory rules, budget related information, and information about open enrollment were some of the contributions identified in the data as sensitive knowledge, beca use any errors during their codification or use may cause monetary, legal, and even reputati onal problems for both employees and organizations. The data collected for this study revealed that ind ividuals were less likely to contribute sensitive knowledge to community-governe d repositories. The major reason is that expert-governed repositories are better equipp ed to maintain the integrity of sensitive knowledge than community-governed repositories. Un like community-governed repositories, expert-governed repositories do not a llow individuals to tamper with contributions through editing. This, in turn, ensu res that the accuracy and integrity of such contributions are not compromised, and do not pose a threat for their future use.
112 This view was supported by one participant who cont ributed general open enrollment information to an expert-governed repository. When the participant was asked if she would have provided the same contribution to the wi ki in her organization, she commented: [This is a] high risk and regulated [information]. If not presented accurately or properly, can cause issues. Would not want others to have the ability to make changes. This perspective was also supported in the second p hase of the study. When participants were asked why they did not provide th eir latest contribution to communitygoverned repositories, 23% of them (or 5 out of 22) chose the pre-coded item that stated I did not want others to edit this contribution. Although this finding could be an artifact of individuals personal preferences (wher e individuals do not want others to edit their contributions for personal reasons rather tha n the sensitivity of contributions), the data provided support (although weak) for the natur e of contributions. At least one of the participants (out of a possible five) who chose the aforementioned pre-coded item contributed sensitive knowledge (about money market s) to the expert-governed repository used in the firm. The remaining contrib utions level of sensitivity could not be evaluated, as participants provided rather general descriptions for the nature of those contributions. Regardless, the data collected from participants suggested that sensitivity was a salient determinant of contribution behaviors for expertand community-governed repositories. This leads to proposing, P11: When expertand community-governed repositori es are used simultaneously, sensitivity of knowledge is (a) pos itively related to contributing to expert-governed repositories, and ( b) negatively related to contributing to community-governed repos itories.
113 It is also worth mentioning that the data provided weak support for an interaction between sensitivity of knowledge and the need for expert validation Since sensitive knowledge may need to be validated by experts, indi viduals likelihood to contribute these types of contributions to expert-governed rep ositories further increases. For example, two of the contributions that were identif ied as sensitive were provided to expert-governed repositories due to the contributor s need for expert validation. Formality of contributions The analysis of the data revealed that formality of contributions played a role in determining which repository individuals chose in p roviding their contributions. For the purposes of this study, formality of a contribution is defined as how well a contribution is structured, or how well it complies with establishe d forms or conventions used in the organization. Accordingly, contributions that have well-defined structures and that comply with established forms, templates, or conven tions can be considered formal; whereas others that convey their message without a certain structure or without complying with a predefined template can be conside red informal. For example, whitepapers, reports, or process documentations can be considered formal contributions; whereas quick and dirty solutions, facts, or enumer ated dos and donts can be considered informal contributions. The analysis of the data revealed that participants were more likely to contribute formal contributions to expert-governed repositorie s, and informal contributions to community-governed repositories. For example, duri ng the first phase of data collection, two participants explicitly mentioned that the cont ributions they were willing to provide
114 were too informal for the expert-governed repository used in their firm. Similarly, another participant explicitly stated that the cont ribution he/she provided for the expertgoverned repository in the firm was too formal for the community-governed repository. Participants in the second phase shared the same co ncern, as 27% of them (or 6 out of 22) chose the pre-coded response that stated expert-go verned repository was to formal for this contribution, and another 27% (or 6 out 22) c hose the pre-coded response that stated community-governed repository was too informal for this contribution. The reason for this finding is the mismatch between the formality of contributions and formality of the governance mechanisms used for repositories in the organization. Expert-governance is considered a more formal mecha nism as it imposes a predefined set of quality standards on submissions by a designated group of experts, whose job is to ensure that all submissions made to the repository comply with these standards. On the other hand, community-governance can be considered a more informal mechanism as community members do not follow stringent quality s tandards to improve the quality of contributions. This, in turn, induces individuals to submit more formal contributions to the expert-governed repositories and more informal ones to the community-governed repositories in firms. This is because the type of the governance mechanism may not be equipped to handle contributions (or increase their quality) if there is a mismatch. For example, unless submissions are well-structured and well-organized, and they comply with the norms and quality standards imposed by the governance mechanism, they can be rejected by expert-governance, or be subjected to g o through several rounds of revisions to make them compatible with existing norms. There fore, informal contributions in the form of unstructured and quick solutions, facts, or best practices not only stand a chance
115 to get published in the repository, but also do not fit well with what is predominantly stored in expert-governed repositories. Similarly, it is not likely for a whitepaper or a report to be contributed to a repository where info rmal contributions such as quick solutions or workarounds to problems are discussed. The interviews provided support for this argument as one participant in the IT industry mentioned that the knowledge assets contributed to the wiki used in the firm were mostl y in the form of notes or bullet points. This was in contrast to the formal contributions in the expert-governed repository that were structurally sound, and followed a standard re port format. Additional support was provided by another participant, who mentioned, [Community-governed repository] is more conversati onal, and [expert-governed repository] is for formal approved communications. Therefore, this study proposes, P12: When expertand community-governed repositori es are used simultaneously, formality of contributions is (a) p ositively related to contributing to expert-governed repositories, an d (b) negatively related to contributing to community-governed repos itories. It is also worth mentioning that a mismatch between the formality of governance mechanism and culture of the organization can influ ence the choice of governance mechanisms. For example, if organizational members are used to exchanging more informal knowledge, an implementation of expert-gov ernance can hinder contributions to repositories as it challenges existing norms in the organization and ultimately cause withdrawal behaviors on the part of organizational members. For example, one participant, who use expert-governed repository in his firm, commented, For the number of staff we have in this organizati on, I don't see much participation. The management tries to encour age people to write some white papers and come up with some plans as a [knowledge sharing] forum. But we still get minima l response.
116 That could be due to the formal [structure] of the process to put these things together. Although the above the comment provides support for a possible interaction between the formality of governance mechanisms and culture, this relationship was not considered in this study due to weak support. Need for collaboration In the context of this study, the term collaboratio n refers to individuals need to co-author contributions, and their desire to get fe edback from others in the organization about their contributions in repositories. The dat a collected for this study revealed that individuals who sought collaboration were more like ly to make contributions to community-governed repositories than expert-governe d repositories. In other words, the need to co-author contributions or get others feed back about a specific contribution encouraged contributing to community-governed repos itories, and discouraged contributing to expert-governed repositories. The data collected in the second phase of the study also supported this finding. When partic ipants were asked why they did not make their latest contribution to the expert-govern ed repository in their firm, 23% (or 5 out of 22) chose the pre-coded response that stated this contribution needed to be coauthored, and 41% (or 9 out of 22) chose the pre-c oded responses that stated that I wanted to get the communitys feedback about this c ontribution. The reason for the above finding is that, by virtue of their design, expert-governed repositories do not support collaboration among org anizational members. They impose restrictions on user-privileges about editing exist ing contributions or providing feedback about them through comments or ratings. This in tu rn creates a major hurdle for
117 individuals who seek others help in codifying new or improving existing contributions in repositories. On the other hand, community-governe d repositories provide a good venue for collaboration, as the technological features af forded by community-governance allow organizational members to co-author with or provide continuous feedback to each other. The support for this argument was provided by one c ustomer support specialist, who mentioned that their work relied heavily on feedbac k among support personnel in resolving customer problems. The community-governe d repository used in the department created a good venue for the department personnel to receive or provide feedback compared to the more static expert-governe d repository. Individuals actively participated in discussions, communicated what work ed and what did not in resolving problems, provided comments about any updates to ex isting solutions, and more importantly, enabled alerts within the system to pu sh these updates to themselves from the repository. On the other hand, they seldom pro vided contributions to the expertgoverned repository, which by design did not allow the department personnel to interact with or provide feedback to each other. The salience of collaboration in choosing a governa nce mechanism was highlighted in another instance during the intervie ws. One participant, who provided her contribution to the community-governed repository u sed in her firm, mentioned, I do not think I would have made that contribution to the expertgoverned repository. The comments [to the contribu tion] only came after it had been in use by several executives Therefore, I do not believe having experts review the [contribut ion] prior to its posting would have helped in any way. In addition to imposing restrictions on co-authorin g and providing feedback, expert-governance also introduces experts as an int ermediate layer between knowledge
118 providers and knowledge seekers, which further stif les collaboration among organizational members. For example, one participa nt commented, My company has a deep knowledge repository. We ar e geographically located across the country and rely heavily on our repository of documents, ideas, toolkits and best p ractices. I believe the layer of experts hinders our informal ity and collaboration. Another participant discussed how the mediating rol e of experts prevented him to make contributions to an expert-governed repository through the following comment, [Expert-governed repository] didnt motivate me to make this contribution. Any contribution that you make should go out as you contribute it. Not altered by some expert. Why shou ld we contribute when any of our comments are altered by an expert. Therefore, the above discussion leads to proposing, P13: When expertand community-governed repositori es are used simultaneously, the need for collaboration is (a) p ositively related to contributing to community-governed repositories, and (b) negatively related to contributing to expert-govern ed repositories. Need for expert validation Expert validation is the process with which contrib utions are vetted by experts before they are published in repositories. Validat ion ensures that contributions are accurate, applicable, reliable, and compliant with the quality standards developed in the organization. The analysis of the data revealed th at if individuals had a need for expert validation for their contributions, they were more likely to provide it to expert-governed repositories than community-governed repositories. The evidence for this finding is provided by several participants. One participant provided his/her contribution to the expert-governed repository used in the firm. When asked if he/she would have made
119 same contribution to the community-governed reposit ory used in the firm, the participant commented, No. [This information] had to be vetted and enhanc ed by the [expert] before getting published. Another participant, who chose the expert-governed repository used in the firm, commented, No, I [wouldnt] feel comfortable if somebody used this information without being validated first. The second phase of the study provided more support for the salience of expert validation. When participants were asked why they did not make their latest contribution to community-governed repositories in their firms, 18% (or 4 out of 22) chose the precoded response that stated this contribution had t o be vetted by experts. Individuals may seek expert validation for two reas ons. The first concerns the type of knowledge being contributed to repositories Although the evidence is rather weak, the analysis of the data suggested that indiv iduals tended to seek expert validation for sensitive knowledge, which is defined in the co ntext of this study as knowledge that can have legal ramifications for individuals or org anizations if not codified or used appropriately. Therefore, if individuals feel that inaccuracies in contributions can get individuals or organizations into financial, legal, or reputational troubles, they may want experts to validate these contributions before they are published in repositories. The second reason is personal preference. Accordin gly, individuals may seek expert validation out of personal preferences regar dless of the type of contribution they provide to repositories. There is support in the d ata for this argument. When one participant was asked if he/she would consider maki ng contributions to communitygoverned repositories (if available in the firm), t he participant commented,
120 I would rather have [experts] to vet all my contri butions Another participant, who used community-governed re pository in her organization, mentioned that one reason why she was not willing to provide contributions was the need for expert validation. She commented, If I could submit something that was maybe 90% pol ished or accurate and have the expert increase the quality, then I'd probably be more inclined to [contribute] to it. One of the reasons for this personal preference can be individuals self-esteem or their confidence in their level of knowledge. A kn owledge management professional of an IT firm, who was responsible for overseeing both the expert-governed repository and the wiki used in the firm, mentioned that expert-go verned repository seemed more attractive for some people due to the availability of expert validation. He mentioned that certain individuals in the organization tended to b e less confident in their level of knowledge and tended to have lower levels of self-e steem compared to others in the organization. This led them to contribute to the e xpert-governed repository instead of the wiki in order to make sure that what their contribu tions were approved by experts before published in the repository. Therefore, this study proposes, P14: When expertand community-governed repositori es are used simultaneously, the need for expert validation is ( a) positively related to contributing to expert-governed reposito ries, and (b) negatively related to contributing to community-gov erned repositories. Need for recognition In an organizational setting, recognition is the ac knowledgement of ones action by supervisors or colleagues (Deci and Ryan, 1985). Prior literature suggests that recognition is an important driver of organizationa l behaviors, as it reinforces individuals
121 to continue performing a behavior (e.g., Amabile, 1 993; Deci and Ryan, 1985). For example, organizational citizenship behaviors, whic h are discretionary behaviors that facilitate the efficient and effective functioning of an organization, are reinforced if they are recognized by others in the organization (McNee ly and Meglino, 1994). Analysis of the data showed that need for recogniti on was a salient determinant of contribution behaviors for expertand community-go verned repositories. Specifically, it had a negative effect for contributing to community -governed repositories, but a positive effect for contributing to expert-governed reposito ries (when these repositories were used simultaneously). The reason for this disparity was that contributions did not get recognized in community-governed repositories. For example, in the first phase of data collection a software engineer mentioned that he wi shed he contributed the documentation of a complex algorithm and its applic ation to the expert-governed repository (instead of the community-governed repos itory), because, [The contribution] would have not only been perfec ted, but would have gained recognition. In another instance, one participant explained why he/she chose expert-governed repository to make a contribution by commenting, So that the contribution was linked to my official personal profile and gained recognition The effect of recognition became more apparent in t he second phase of the study. When individuals were asked why they did not provid e their latest contribution to community-governed repositories in their firm, 14% (or 4 out of 29) chose the pre-coded response which stated that contributions do not ge t recognition in the communitygoverned repository. Further, one participant of the second phase commented,
122 I like making contributions to our wiki. It gives me a great deal of satisfaction. However, if you make a contributi on to the wiki to, say, gain recognition, you are [going to] walk away empty handed. The above evidence suggests that community-governed repositories are not a good venue for individuals to gain recognition for their contributions. For this reason, individuals who strive for gaining recognition in t heir organizations will be less likely to contribute to community-governed repositories if th ere is also an expert-governed repository in those organizations. There can be mu ltiple reasons why contributions made to community-governed repositories are not recogniz ed. First, knowledge users may not easily identify the contributor of a knowledge asse t provided to these repositories. For example, multiple individuals can contribute to a w iki page, making it difficult to identify the contribution of a single individual. This, in turn, makes it harder for others in the organization to recognize contribution efforts or g ive those individuals credit. Similarly, in a discussion forum, a thread itself may become a valuable piece of knowledge in its entirety, while it may be difficult, if not impossi ble, to recognize the efforts of all the contributors in that thread. Therefore, communitygoverned repositories reliance on collective effort may hinder acknowledging each ind ividuals effort, which may discourage contribution behaviors. Two of the abov e comments highlight this problem, as participants suggest that community-governed rep ositories are not a good venue to make the association between the contributor and th e contribution. Second, community-governed repositories are usually implemented as an experimental technology in most organizations. The refore, contributions provided to these repositories are perceived as discretionary e fforts that result from individuals own interest or enthusiasm for using of those technolog ies. This, in turn, eliminates the
123 possibility of supervisors or peers to evaluate the se discretionary efforts using the formal and informal reward structures employed in organiza tions. This argument received support from one of the participants in the study. When the participant was asked if the contribution could have been made to the communitygoverned instead of the expertgoverned repository, he/she commented, NO. Because there are no appropriate incentive sys tems in place to reward the effort. Regardless of the reason, participants of the study consistently mentioned that their inability to gain recognition from contributi ons made to community-governed repositories induced them to provide contributions to expert-governed repositories. Therefore, this study proposes, P15: When expertand community-governed repositori es are used simultaneously, the need for recognition is (a) pos itively related to contributing to expert-governed repositories, and ( b) negatively related to contributing to community-governed repos itories. Trustworthiness of Findings This essay uses the approach proposed by Lincoln an d Guba (1985) to assess the trustworthiness of findings. In an effort to reduc e duplication, readers can refer to the Trustworthiness of Findings section of the first es say for more information about the criteria used for trustworthiness, and how the find ings rate against these criteria. Discussion Key findings The research question of interest to this study was : what factors influence individuals to make voluntary contributions to expe rtand community-governed repositories. This research question was examined in two different contexts, one in
124 which there is only one type of repository (either an expert-governed or a communitygoverned), and another in which the two types of re positories exist simultaneously. The analysis in the first context intended to compare t he factors that were salient for making contributions to expert-governed repositories with those for making contributions to community-governed repositories. On the other hand the analysis in the second context intended to understand the factors that induced ind ividuals to choose one type of repository over another. The analyses in both cont exts were conducted using an interpretive paradigm. Qualitative research was co nducted using grounded theory to uncover the salient factors from empirical data tha t were collected from organizational members in various organizations using interviews a nd online questionnaires. The findings suggested that when organizations empl oyed only one type of repository, the factors that explained contribution behaviors in expert-governed repositories were similar to those in community-gov erned repositories. Specifically, organizational benefits, reputation, altruism, and organizational rewards were positively related to making contribution to both types of rep ositories; and codification effort, lack of expertise and risk of duplication were negativel y related to contributing to both repositories. The two differences between the moti vating factors of the repositories were that personal benefits positively influenced contri bution behaviors only for expertgoverned repositories, and reciprocity positively a ffected contribution behaviors only for community-governed repositories. These findings suggest that, when organizations use one type of repository (either expertor community-governed), a general set of fa ctors can adequately explain employees contributions behaviors. However, expla natory power can be increased when
125 personal benefits (for expert-governed repositories ), and reciprocity (for communitygoverned repositories) are also considered. Person al benefits is related more to individuals self development, as individuals contr ibute with the expectations that those contributions will increase their own efficiency an d effectiveness, and thus help them perform better in their jobs. On the other hand, r eciprocity is more related to the concept of social exchange, as contributions are provided t o fulfill obligations from previously received help, or to get help in the future from ot hers when needed. The second set of findings is for the context in wh ich the two types of repositories are used simultaneously. The analysis showed that when organizations used both expertand community-governed repositories side-by-side tw o sets of factors explained contribution behaviors: knowledge-based and need-ba sed. Knowledge-based factors concerned whether contributions were suggestions/id eas, and to what extent they were sensitive and formal. Accordingly, suggestions/ide as, non-sensitive contributions, and informal contributions were more likely to be contr ibuted to community-governed repositories; and sensitive, and formal contributio ns were more likely to be contributed to expert-governed repositories. These findings indic ate that the type as well as the characteristics of knowledge play a role in explain ing which repository individuals are more likely to contribute to. On the other hand, need-based factors concerned col laboration, expert-validation, and recognition. Findings revealed that individual s who needed to collaborate (for example, to co-author contributions) were more like ly to choose community-governed repositories for their contributions, whereas indiv iduals who needed expert-validation or
126 recognition for their contributions were more likel y to contribute to expert-governed repositories. The above findings show that the sets of factors th at influence contribution behaviors greatly differ in the two contexts in whi ch the research question is investigated. The reason for this difference is that in the first context (where there is either an expertor a community-governed repository), the decision t o make a contribution suppresses the salience of the governance mechanism used for the r epository. When asked about their motivations to contribute, participants focused on the act of contribution rather than the governance mechanism or the contributions fit for the mechanism. In the second context, however, the choice of repository for prov iding a contribution is more important than the act of contribution. In this context, the decision to make a contribution has already been made and the focus is on choosing the right repository or the governance mechanism for the contribution. Therefore, when as ked about their motivations to make contributions, participants focused on the contribu tions fit for the governance mechanism rather than their motivations to codify c ontributions. Limitations of the study This study is not without its limitations. Since t he data for this essay and the first essay were collected at the same time, the limitati ons of the first essay apply to this essay as well. Therefore, readers can refer to the Limit ations section of the first essay to see the pitfalls of this study and understand how the findi ngs need to be interpreted. An additional limitation of this essay is that the study did not distinguish between voluntary and mandatory contribution behaviors duri ng data collection. The data collection instruments did not ask participants whe ther providing contributions to
127 repositories in their organizations were a requirem ent of their job descriptions or were voluntary behaviors. For this reason, some of the participants mentioned that it was their duty to make a contribution when they were asked ab out their motivations. Although this does not necessarily pose a threat to the internal validity of the findings, it reduces the amount of usable answers for constructing the first theoretical model proposed in this essay. However, the responses of these individuals are still valuable for the second theoretical model, which is relatively robust with respect to the mandated contribution behaviors as the model focuses on choosing one type of repository over another in making a contribution rather than the decision to m ake the contribution. Theoretical implications This study has several theoretical implications. F irst, this study is one of the earliest studies that differentiate the factors tha t motivate individuals to contribute to expert-governed repositories from those that motiva te them to contribute to communitygoverned repositories. Prior research does not tak e the governance mechanisms into account in explaining contribution behaviors, and t herefore, implicitly assumes that individuals are motivated in the same way for provi ding contributions to both types of repositories. This study challenges this view, and suggests that new theoretical understandings need to be developed for explaining contribution behaviors to repositories that are governed by different mechanisms. Therefo re, this study develops two different theoretical models using grounded theory approach t o explain individuals motivations to contribute to expertand community-governed reposi tories. Although certain factors in these two models overlap, the differences that stem from the concepts of selfdevelopment and social exchange are worthy of theor etical consideration. This study
128 contributes to our existing theoretical knowledge b ase by suggesting that future theory development efforts should focus on the self-develo pment concept for expert-governed repositories, and the social exchange concept for c ommunity-governed repositories to have a deeper understanding of contribution behavio rs. The second theoretical implication of this study is that researchers may need to focus on explaining contribution behaviors (for eit her type of repository) by differentiation between the enabling and the inhibi ting factors. In this study, the enabling factors are those that are positively related to co ntributing to both repositories, and the inhibiting factors are those that are negatively re lated to contributing to both. The benefit of making this distinction is that the explanatory power of theories that explain contribution behaviors can be increased by theorizi ng the effect of each construct appropriately. For instance, enabling factors oper ate along the positive and negative spectrum, explaining why individuals make or fail t o make contributions to repositories; whereas inhibiting factors operate only in the nega tive spectrum, as they do not have a meaningful opposite or their opposites do not have a positive effect. For example, organizational rewards is considered an enabling fa ctor, because it can affect contribution behaviors both positively and negatively (i.e., mor e rewards can increase contribution behaviors, and less rewards can reduce them). On t he other hand, lack of expertise is considered an inhibiting factor, as it only explain s why individuals fail to make contributions to repositories. Theorizing a positi ve relationship between expertise and contributions can lead to non-significant findings (c.f., Wasko and Faraj, 2005), as expertise is a necessary but not a sufficient condi tion for making contributions, and therefore, does not necessarily induce more contrib utions. The concept of enablers and
129 inhibitors is rooted in the work of Centefelli (200 4), which suggests that social behaviors can be explained using two different sets of constr ucts that have differing variability along the positive-negative spectrum. In fact, thi s view is no stranger to the management literature, as Herzberg et al.s (1959) work on mot ivation-hygiene theory in the area of organizational behavior advocates that certain jobrelated factors increase employee satisfaction (and therefore act as motivators); whereas another set of factors increase dissatisfaction (and therefore act as de-motivators). They argue that factors that cause dissatisfaction do not operate in the reverse direc tion (as company policies may cause employee dissatisfaction; but may not necessarily c ontribute to satisfaction). There is a need to apply the same principle in explaining cont ribution behaviors, as prior literature does not differentiate between enabling and inhibit ing factors. Such a distinction may not only increase explanatory power of theories that ex plain contribution behaviors, but also reconcile the conflicting findings in the literatur e. The third, and the final, theoretical implication o f this study is that this study develops a theoretical model that explains the cond itions under which individuals choose expert-governed repositories over community-governe d repositories (and vice versa) to provide their contributions, when these two reposit ories are implemented simultaneously. The findings argue that if organizations implement these two types of repositories sideby-side, individuals make deliberate, instead of ra ndom, choices in contributing to one repository over another. The choice behavior is ex plained using two different sets of factors, knowledge-based and need-based. Given tha t prior studies in the literature do not distinguish between different governance mechanisms these two sets of factors are
130 expected to pave the way for the development of new theories especially in contexts where there are alternative repositories with diffe rent governance mechanisms. This study has important implications for future re search as well. First, the theoretical relationships proposed in this study co ntribute to the efforts for the development of a theory of contribution. Future st udies that intend to develop a unifying theoretical framework for explaining contribution b ehaviors can, first, test the relationships proposed in this study using a positi vistic perspective, and then incorporate them into previous findings in the literature. To do this, future research can include governance mechanisms as a contingent or a moderati ng variable into existing frameworks, and theorize the corresponding relation ships drawing upon the findings reported in this study. Second, future studies can delve deeper into the tw o factors, personal benefits and reciprocity, which were identified as motivators of contribution behaviors for expertand community-governed repositories respectively. Rese archers can start by studying whether these two factors emanate from organization al culture irrespective of governance mechanisms, or are a consequence of the use of the related governance mechanisms. This is important, because the former suggests that an organization should implement a specific governance mechanism depending on how well the mechanism fits the culture of the organization; whereas the latter suggests that governance mechanisms are effective change agents and have the ability to influence org anizational culture. Further, future research can focus on theories that emphasize selfdevelopment or personal improvement to have a deeper understanding of contribution beha viors in expert-governed repositories;
131 and theories that explain social exchange and recip rocity to uncover the intricacies of contribution behaviors in community-governed reposi tories. Third, researchers can further the theoretical mode ls proposed in this study by incorporating contingent variables that are not con sidered in this study. Incorporating contingent variables is important, since it may inc rease the generalizability of the theoretical models to different contexts. For exam ple, future research can look into the dynamism of the environment, and study whether indi viduals contribution or choice behaviors differ in dynamic (or turbulent) environm ents compared to more stable (or static) environments. In dynamic environments, cer tain types of knowledge may only be valuable if published and used immediately. Anecdo tal evidence suggests that community-governance may be more appropriate for th ese environments since knowledge assets are not exposed to pre-publication processes as in expert-governance. The vetting processes employed by experts can lengt hen the publication time, which may cause the knowledge asset to lose its value. Altho ugh community-governance may be a better alternative in these environments, future re search should investigate the trade-offs between the two governance mechanisms and try to ex amine other variables such as sensitivity of knowledge that may influence contr ibution and choice behaviors in those contexts. Practical implications This study has important implications for practitio ners as well. First, practitioners can use the findings reported in this study to fost er contribution behaviors for expertand community-governed repositories. Especially the fi ndings concerning organizational rewards, organizational benefits, and reputation su ggest that a variety of tools can be
132 leveraged to increase the number of contributions p rovided to both expert-governed and community-governed repositories. For example, prac titioners can create formal policies that define rewards, or develop incentives to align employees goals with those of organizations. To further increase contribution be haviors, practitioners can adopt more targeted approaches for each type of repository. F or expert-governed repositories, they can promote the benefits of contributions for perso nal development. For example, the benefits of expert-governed repositories in cleansi ng, standardizing, organizing, and storing personal knowledge can be communicated to e mployees and the benefits of such knowledge on individuals performance evaluations c an be emphasized. For communitygoverned repositories, practitioners can promote re ciprocity by increasing the transparency of knowledge contribution and knowledg e use processes in the organization. Specifically, meta-data about the number of contrib utions made to repositories versus the extent of knowledge used from repositories can be c ommunicated in an effort to stimulate a sense of fairness among organizational members. Second, this study informs practitioners of the con ditions under which employees prefer expert-governed repositories over communitygoverned repositories (and vice versa) when these two repositories are implemented simultaneously. Given that most organizations are starting to use these two types o f repositories side-by-side to provide more opportunities to employees for sharing knowled ge, the findings reported in this study can be used to understand why employees tend to use one repository but not the other. Drawing upon the findings, practitioners ca n assess the types and the characteristics of knowledge being shared in an org anizational unit, determine the needs (or predispositions) of employees in that unit, and make more informed decisions about
133 what type of repository to implement for sharing kn owledge in the unit. This is important as organizations can save resources by implementing the type of repository that best serves the needs of individuals in a given organiza tional unit, and by avoiding the implementation of an alternative repository that ma y be less likely to be used by organizational members. Third, findings suggest that expertand communitygoverned repositories may not be substitutes of each other in organizational settings. Employees use these repositories to share different types of knowledge and to satisfy different types of needs. For example, individuals are not likely to share fo rmal and sensitive knowledge in community-governed repositories, and informal knowl edge or suggestions/ideas in expert-governed repositories. Therefore, if organi zations want to cater different needs of individuals, or want their employees to share diffe rent types of knowledge, they may need to consider implementing both expertand comm unity-governed repositories, and promote these repositories appropriately in the org anization.
134 ESSAY III: THE ROLE OF GOVERNANCE MECHANISMS IN USI NG KNOWLEDGE FROM REPOSITORIES Introduction The use of expertand community-governed repositor ies as a means to facilitate knowledge transfer among individuals is increasing. However, the current literature neither distinguishes between these two types of re positories, nor examines the factors that affect individuals use of knowledge from thes e repositories. Therefore, the goal of this essay is to understand the factors that influe nce individuals use of knowledge from expertand community-governed repositories, where knowledge use, in the context of this study, is defined as retrieving explicit knowl edge from electronic repositories and employing it to perform a task (Nonaka, 1994). The research questions of interest to this essay are: (a) what factors influence individuals use of knowledge from expertand community-governed repositories; and (b) how? The motivation to examine these research questions is rooted in our limited understanding of knowledge use from repositories th at are governed with different mechanisms. Prior research does not shed any light on different forms of governance, and therefore, provides no guidance on how knowledg e use behaviors may differ in the existence of governance mechanisms. Given the prev alence of governance mechanisms, there is a need to understand whether our existing theoretical understanding of knowledge use needs to be revised to explain knowledge use fr om repositories with governance
135 mechanisms, and if yes, how. This essay attempts t o address this gap in the literature through a positivist paradigm by drawing upon elabo ration likelihood model (ELM) from social psychology(Petty and Cacioppo, 1986a; Petty and Cacioppo, 1986b). This essay is organized as follows. In the next se ction prior literature on knowledge use is reviewed, and gaps in the literatu re are identified about using knowledge from repositories with governance mechani sms. In the following section, the theoretical framework used in this study (i.e., ELM ) is discussed, and research hypotheses are formulated. The next section presen ts the research methods used in this study, which discusses the details of the experimen tal design employed for this essay. In the following section, the findings of the study ar e presented, followed by the discussion section, which summarizes the key findings, limitat ions, and theoretical and practical implications of this study. Prior Research This essay defines knowledge use as retrieving explicit knowledge from electronic repositories and employing it to perform a task (Nonaka, 1994). This definition is in line with the existing literature although researchers u se different terminology to refer to knowledge use. For example, knowledge use, as defi ned in this essay, is also referred to as knowledge reuse (e.g., Markus, 2001), knowledge adoption (e.g., Sussman and Siegal, 2003), knowledge utilization (e.g., Larsen, 1980), and knowledge application (e.g., Alavi and Leidner, 2001; Holzner and Marx, 1979; Wiig, 19 95) in the literature. Despite the terminological differences, all conceptualizations involve retrieval or transfer of knowledge, and leveraging this knowledge to perform certain tasks.
136 The challenge in studying knowledge use is not the existence of different terminology, but the underlying cognitive process. Since, our knowledge of what really transpires in the minds of individuals during knowl edge use is still limited (Sussman and Siegal, 2003), empirical investigations are inheren tly challenging. Prior attempts to conceptualize these cognitive processes suggest tha t individuals may use knowledge at a low or a high level (Caplan, 1975; Rich, 1975). Th e low-level knowledge use has a very narrow scope and involves performing the set of act ions prescribed by a knowledge asset. For instance, configuring an e-mail client such as Microsoft Outlook to retrieve e-mail messages from the firms e-mail server can be consi dered a low-level knowledge use, because individuals need to follow a set of instruc tions verbatim for successful configuration. Studying this type of knowledge use may not be as problematic, since individuals are expected to perform the set of acti ons exactly they are prescribed. However, high-level knowledge use is much broader, and involves an enlightenment process (Weiss, 1979). In this case, individuals m ay not necessarily perform the specific actions prescribed by the knowledge, but may blend it with what they already know to perform an adapted, a reinvented, or a modified act ion. In this case, the performed action may not mimic what the original knowledge prescribe s. This type of knowledge use is more prevalent in policy-making, as Caplan (1975) s hows that nearly 10% of actions taken by 204 upper-level executives in the US gover nment can be characterized as metalevel knowledge use. This suggests that caution ne eds to be taken in studying knowledge use, as knowledge use may not necessarily mean that individuals will perform the actions exactly as they are described (Larsen, 1980; Oh, 19 97).
137 Prior literature investigates the antecedents of kn owledge use in two separate streams. The first stream adopts a macro view and treats knowledge use as an overarching concept and a set of processes consisti ng of capturing, packaging, distribution, and application of knowledge. The go al of this stream is to identify the conditions under which these processes are facilita ted. For example, Dixon (2000) suggests five types of knowledge use (i.e., transfe r) situations, namely serial, near, far, strategic, and expert transfer, contingent upon who the receiver is, what type of task is being performed, and what type of knowledge is bein g transferred. Similarly, Markus (2001) suggests four types of knowledge use situati ons by focusing on who the users of knowledge are: shared work producers, shared work p ractitioners, expertise-seeking novices, and secondary knowledge miners). This str eam provides two important insights: (1) individuals are less likely to use knowledge if the conditions that define a knowledge use situation are not met; and (2) knowledge reposi tories can support knowledge transfer by storing high-quality, de-contextualized, and eas y-to-understand knowledge, and by providing certain design features such as indexing and search capabilities. The second stream of research adopts a narrower vie w and investigates knowledge use by focusing on whether individuals ad opt knowledge stored in repositories. The dominant theoretical framework u sed in this stream is elaboration likelihood model (ELM; Petty and Cacioppo, 1986a). Since ELM was originally developed to study persuasion and attitude change, studies in this stream extend ELM to the KM context and suggest that knowledge use occur s if individuals perceive its quality to be high, and its source to be credible contingen t upon knowledge users elaboration likelihood (i.e., expertise and involvement in the subject matter). Among the studies that
138 examine this perspective, Mak et al. (1997) conduct an experiment to investigate users acceptance of an expert systems recommendations. They use ambiguity of the decision setting and credibility of experts to understand in dividuals acceptance of recommendations. They use users participation in the design of the expert system as a proxy to the elaboration likelihood motivation. Th eir findings parallel ELMs predictions such that users who participate more in the design accept recommendations if the decision setting is ambiguous, and users who partic ipate less in the design accept recommendations if these recommendations are provid ed by credible experts. Dijkstra and colleagues (Dijkstra, 1995; Dijkstra, 1999; Dijkstra et al., 1998) conduct three experiments in order to study the per suasiveness of expert systems. In the first experiment (Dijkstra, 1995), subjects, unexpe ctedly, rely on credibility of the system rather than the argument quality even though they h ave prior expertise in the subject matter. This leads to conducting the second experi ment (Dijkstra et al., 1998), which suggests that subjects perceive the expert system m ore persuasive than humans even though both sources give the same advice. The resu lts also show that elaboration likelihood of individuals do not matter in determin ing the persuasiveness of the expert system. Finally, third experiment (Dijkstra, 1999) investigates why subjects agree with incorrect advice provided by the expert system, and reports that subjects who tend to disagree with the advice engage in critical thinkin g, while subjects who agree with incorrect advice rely more on cues. Sussman and Siegal (2003) investigate the likelihoo d that consultants at a public accounting firm adopt information provided in elect ronic mail. Unlike ELMs original dependent variable (i.e., attitude change), they us e consultants beliefs about information,
139 which is operationalized using perceived usefulness of information. In line with the predictions of ELM, they report that argument quali ty and source credibility are positively related to consultants perceived useful ness of information, which, in turn, leads to adoption of information provided in the em ails. The moderating effects of elaboration likelihood also conform to the theorys predictions as consultants rely more on the central route if their expertise and involve ment in the subject matter are high. Similarly, Fadel et al. (2008) investigate whether perceived usefulness of information leads to information adoption using an experiment that uses a mock knowledge repository and recommends Internet authen tication solutions. In addition to the constructs of ELM, they add another peripheral route construct to account for the validation of knowledge in repositories. While the y fail to support ELMs predictions they suggest that validation of information is posi tively related to its perceived usefulness. The second stream also includes non-ELM studies tha t draw upon different theories. For example, Zhang and Watts (2008) use Heuristic-Systematic Model (HSM; Chaiken, 1980; Chaiken et al., 1989) as an alternat ive dual-process model to investigate how individuals adopt information from online commu nities. Similar to ELM, they operationalize systematic processing using argument quality, and heuristic processing using source credibility, both of which are moderat ed by disconfirming information and focused search in order to account for the attenuat ion tenet of HSM. Studying two discussion forums, they support argument quality an d source credibility as determinants of information adoption, but provide mixed support for the moderating impacts of disconfirming information and focused search.
140 The second stream provides us with two key insights First, individuals are more likely to use knowledge if they find the knowledge to be of high quality and the source to be credible. Second, argument quality and source c redibility have varying effects on knowledge use contingent on individuals ability an d motivation to elaborate, which are their expertise and involvement in the subject matt er respectively. However, the existing literature overlooks governance mechanisms, and doe s not consider the possible effects of how governance mechanisms influence knowledge us e. The cross-tabulation of the current literature with respect to the type of repo sitory and governance mechanism, as presented in Table 17, reveals that the majority of studies examine knowledge use in expert-governed organizational repositories. Type of repository Governance mechanism Organizational Non-organizational Expert-governance Dijkstra (1995) Mak et al. (1997) Dijkstra et al. (1998) Dijkstra (1999) Fadel et al. (2008) Community-governance ( ) Zhang and Watts (2008) No governance Sussman and Siegal (2003) Table 17. Types of repositories studied and their g overnance mechanisms Of the studies surveyed in the literature, only Zha ng and Watts (2008) focus on knowledge use from a community-governed repository without specifically referring to it. However, none of these studies take the govern ance mechanisms used for the examined repositories into account. This suggests that there is a gap in the literature
141 about the possible effects of governance mechanisms especially expertand communitygovernance, on knowledge use. This study addresses this gap by proposing a research model rooted in elaboration likelihood model of soc ial psychology. Theory and Research Model Elaboration Likelihood Model This essay uses elaboration likelihood model (ELM; Petty and Cacioppo, 1986a; Petty and Cacioppo, 1986b) as a dual-process theory to study the research questions of interest. ELM is appropriate for the purposes of t his essay, because it explains how individuals form attitudes toward objects, issues, or people (Petty and Cacioppo, 1986a). Since the problem of knowledge use can be represent ed as a problem of attitude formation toward knowledge assets, ELM can provide insights about explaining knowledge use from electronic repositories. In fac t, the problem of knowledge use has already been represented as the problem of attitude formation by numerous studies in KM literature. For example, ELM (and its variants) ha s been used to understand whether employees are persuaded by suggestions provided by expert systems (e.g., Dijkstra et al., 1998; Mak et al., 1997), whether employees adopt kn owledge provided by their colleagues (Sussman and Siegal, 2003), or whether i ndividuals adopt knowledge provided in web-based online communities (Zhang and Watts, 2008). In explaining how individuals form attitudes toward objects, issues, or people, ELM draws upon the dual-process perspective rooted in social psychology. It suggests that two alternative processes (hereafter referred to as routes) contribute to attitude formation: central and peripheral routes. In the c entral route, individuals scrutinize the merits or demerits of available information about t he object or argument before forming
142 an informed judgment. They form strong attitudes i f they perceive the information as being of high quality. This process, called elaboration is time-consuming, demanding, and effortful on the part of knowledge users. In t he peripheral route, on the other hand, individuals rely on cues, such as credibility of th e information source, in forming attitudes toward objects or arguments. In this cas e favorable attitudes form not because of the merits of an argument, but because the argum ent comes from a credible knowledge source. This route requires less cognitive effort, is fast and automatic, and does not involve elaboration. The central and peripheral ro utes are commonly operationalized in ELM using argument quality and source credibility c onstructs. Argument quality refers to the users perception of the validity, appropria teness, and accuracy of the argument presented in regards to the attitude object, while source credibility refers to their perceptions of the expertise and trustworthiness of the argument source (Pornpitakpan, 2004). ELM suggests that a contingent factor, called elaboration likelihood determines whether individuals invoke the central or the perip heral route to form attitudes. Elaboration likelihood refers to individuals ability and motiv ation to elaborate, and is predominantly operationalized using individuals ex pertise and involvement (respectively) in the subject matter. Individuals with high elaboration likelihood are more likely to employ the central route, since they are more capable of managing the cognitive effort involved in evaluating the merits of an argument. On the other hand, individuals with low elaboration likelihood are mor e likely to employ the peripheral route, as they lack the ability and motivation to e laborate, and therefore attend to cues such as source credibility to form judgments.
143 Subsequent ELM research suggests that central and p eripheral routes may not work in isolation but may impact one another. For instance, Slater and Rouner (1996) suggest that it is possible for individuals to eval uate the quality of an argument from the credibility of its source and vice versa. This arg ument is consistent with dual process theorists suggestion that individuals have an inna te desire to achieve congruency between the responses generated by central and peri pheral routes (Festinger, 1957; Gawronski and Bodenhausen, 2006; Sloman, 1996). In congruent responses create cognitive discomfort, which may lead individuals to update one of the responses to make it compatible with the other. For example, individ uals facing two conflicting responses about an argument (e.g., the source is credible but the argument is of low quality) can justify their favorable attitudes toward that argum ent by making themselves believe that the argument should be of high quality since it com es from a credible source (or that the source should be less credible than initially thoug ht). In this case, individuals rationalize their decision by updating the response generated b y one of the routes. Research Model To apply ELM to this studys context, its dependent variable needs to be extended to explain using knowledge from repositories. Give n its focus on attitude formation, ELM employs attitude as the primary dependent varia ble of interest. However, prior research on attitude formation suggest that individ uals attitudes toward an attitude object are manifested in their intentions regarding that o bject, which subsequently influences their behavior regarding that object (e.g., Petty e t al., 1983). Although some researchers (e.g., Ajzen and Fishbein, 1980) draw a distinction between attitude and intention, technology acceptance research (e.g., Venkatesh et al., 2003) views attitudes as being
144 embedded in and redundant with intentions. Consist ent with the later stream of research, attitude is represented as individuals intention t o use that knowledge asset, which is purported to influence knowledge use in a positive manner. This expectation, illustrated in the research model in Figure 9, leads to the fir st hypothesis: H1: Users intention to use (a) expert-governed or (b) communitygoverned knowledge assets is positively related to their actual usage of those knowledge assets. Description of constructs: Quality : Quality of knowledge asset; Credibility of gov mech .. : Credibility of the governance mechanism in place; Credibility of source : Credibility of the source of knowledge; Elaboration : Individuals ability and motivation to elaborate (operationalized as user expertise and user involvement); Intention : Intention to use the knowledge asset; Knowledge use : Use of the knowledge asset Figure 9. Research Model for Essay III Based on ELM, it is inferred that ones attitude to ward a knowledge asset is determined jointly by his/her perceptions of the qu ality of that knowledge (the central route) and the credibility of the knowledge source (the peripheral route). If individuals perceive the knowledge asset as being high-quality, theyll have favorable attitudes
145 toward that knowledge regardless of the type of gov ernance mechanism used in the knowledge repository. Likewise, knowledge coming f rom a credible source is more likely to induce favorable attitudes among individu als than knowledge coming from less credible sources, regardless of the type of governa nce mechanism used in the repository. The positive associations between source credibilit y, knowledge quality, and intention to use knowledge, as suggested by ELM, are shown in Fi gure 9. However, these associations are not stated as formal hypotheses si nce they are not new in knowledge use research. The presence of governance mechanisms introduces an additional peripheral cue, the credibility of the governance mechanism referring to individuals perceptions of the trustworthiness and reliability of both the governo rs, and the specific page in the repository as a result of the governance processes. If individuals find governance mechanisms credible, they can still have positive a ttitudes toward this knowledge, even if they have little information about the credibility of the knowledge source or are unable to adequately assess knowledge quality. In contrast, if they do not perceive the governance mechanisms as being credible, this perception can u ndermine their attitude toward knowledge derived from these repositories. Therefo re: H2: Credibility of (a) expert-governance or (b) com munitygovernance is positively related to intention to us e knowledge assets. As discussed earlier, the central and peripheral ro utes to attitude formation may be moderated by the elaboration likelihood of knowl edge users. Individuals possessing the motivation and ability to elaborate tend to rel y more on central route and carefully scrutinize the merits or demerits of knowledge asse ts (i.e., argument quality); whereas if they lack elaboration motivation or ability, they m ust rely on peripheral cues such as
146 credibility of knowledge source or of the governanc e mechanism. It should be noted that elaboration is not a personality trait, but rather a situational state that depends on the subjects prior expertise of and exposure to the at titude object. For instance, a physician may elaborate medical arguments because such argume nts are related to his/her profession and he/she has the ability to process su ch arguments, but not elaborate arguments about automotive repair when his/her car breaks down. Drawing from this example, elaboration motivation and ability is conc eptualized as user involvement and user expertise respectively. User involvement and expertise ofte n tend to be positively correlated, but not necessarily so, because a novic e knowledge worker may be deeply involved in a task context, yet lack the expertise of a senior worker in understanding the complexities of that task. Knowledge users with hi gh involvement and high expertise tend to develop more favorable attitudes toward kno wledge assets when presented with high quality arguments, while those with low involv ement and low expertise have more favorable attitudes when presented with a highly cr edible source or a governance mechanism of high credibility. These associations, or in other words the moderating effects of the elaboration likelihood, are not hypo thesized in the research model, as users elaboration likelihood (i.e., their expertise and i nvolvement) is controlled in this study. Therefore, these associations are depicted as dashe d lines in the research model, indicating that their effects are not tested. Although ELM states that central and peripheral rou tes work independently, subsequent studies have suggested that these routes may influence each other. Slater and Rouner (1996) argue that knowledge coming from a cr edible source may be viewed as being high quality argument. Conversely, an unknow n source can be viewed as being
147 credible if arguments provided by this source are d eemed to be of high quality. However, in any given instance, peripheral cues are more lik ely to influence the central route rather than vice versa. This is because peripheral route relies on a slow-learning system in which associating a response with a particular cue requires individuals to be repeatedly exposed to that cue over an extended period of time (Smith and DeCoster, 2000). For example, individual A can perceive individual B as credible only after A interacts with B numerous times. Once created, such perception is s table and unlikely to change unless something occurs to engender a change. In this cas e, A will not likely change his/her perceptions of B with every interaction, because do ing so will impose a significant information processing load on A and can also cause cognitive dissonance due to the temporal instability of knowledge (Smith and DeCost er, 2000). For this reason, central route processing is less likely to influence periph eral cues, as any such possible impact will be spread across time. Hence, credibility of source and the governance mechanism should influence knowledge quality, rather than the reverse, at any given instant of time. However, this study only hypothesizes the effect of credibility of governance mechanism on knowledge quality. Therefore,: H3: Credibility of governance mechanism is positive ly related to the quality of (a) expert-governed or (b) community -governed knowledge assets. Research Methods Subjects and Design The proposed hypotheses were tested using an experi ment at a university located in the southeast US. Subjects were undergraduate b usiness students enrolled in three
148 different courses in the Management Information Sys tems (MIS) program. Participation was voluntary and students received extra credit fo r taking part in the experiment. The goal of the experiment was to provide subjects with two Web pages one expert-governed and one community-governed and un derstand how they used knowledge from these pages to perform an experiment al task. In order to test the hypotheses, the credibility of the governance mecha nisms of both pages were manipulated by setting them either to a high or to a low credibility condition. The manipulation was performed in two ways. The fi rst involved visual cues presented at the top of each page about the gover nance mechanism used for that page. For the expert-governed page, these cues included t he submitter name, the reviewer name, the number of revisions, the submission date, and the publication date at the top of the page. For the community-governed page, the cue s included the submitter name, the number of edits, the number of unique editors, the last edit date, and the rating provided by community members. Second, subjects were given a brief description of these visual cues, which provided the details of the governance functions and the credibility of the individuals who performed these functions. In expe rt-governance, subjects were given details about the credibility of the expert (who re viewed the submissions), and the governance functions employed by the expert; wherea s in community-governance, subjects were given details about the credibility o f the community, and the governance functions performed by the community. Appendix A p resents all four pages used in the experiment (including the visual cues described abo ve), and shows the way with which the visual were described and presented to subjects
149 The high and low credibility conditions for both ex pertand community-governed pages resulted in four different groups, as present ed in Figure 10, where each group was given one expert-governed page and one community-go verned page with different levels of credibility. For example, subjects in the first group were given one expert-governed and one community-governed page, where both pages w ere set to the high credibility condition. Similarly, subjects in the second group were given the same two pages, but the expert-governed page was set to the high credib ility condition, and the communitygoverned page was set to the low credibility condit ion. The cross-product of the rest of the credibility conditions resulted in the third an d fourth groups presented in Figure 10. Observation 1 Treatment 1 Observation 2 Treatment 2 Observation 3 Group 1 O1 EG-H O2 CG-H O3 Group 2 O1 EG-H O2 CG-L O3 Group 3 O1 EG-L O2 CG-H O3 Group 4 O1 EG-L O2 CG-L O3 Legend : EG-H : high credibility expert-governed Web page; EH-L : low credibility expert-governed Web page; CG-H : high credibility community-governed Web page; CG-L : low credibility community-governed Web page; O1 : initial measurement on subjects expertise and in volvement; O2 : measurement on Treatment 1; O3 : measurement on Treatment 2 Figure 10. Experimental design Since measurements were taken from each subject for two pages, the experimental design resembled a repeated measures d esign with the exception that the measures were for different Web pages. Overall, th ree sets of measurements were taken from subjects in the order shown in Figure 10. The first measurement (O1) concerned subjects expertise and involvement in the subject matter to determine whether subject s were familiar with the experimental task or not. T he second and the third measurements (O2 and O3 respectively) concerned subjects percep tions of the first and the second Web
150 pages used in each group respectively. In order to determine if the order of the Web pages influenced subjects perceptions, a counterbalanced design was employed (Grant, 1948; Pollatsek and Well, 1995). Therefore, four a dditional groups were created by reversing the order of the Web pages in Figure 10. This resulted in a total of eight distinct groups: four of which were given the pages in the order presented in Figure 10, and the remaining four were given the pages in the reverse order. Experimental setup The measurement instrument was developed using the services of a popular vendor on the Web that offered online questionnaire s. Using the template questions provided by the vendor, a total of four different m easurement instruments were created for the four experimental groups. Although the sam e measurement items were used for all instruments, four different instruments had to be created to accommodate the different types of treatments used in each experimental group The instruments were hosted on vendors Web servers, and were accessible using the Web link provided by the vendor. Each instrument consisted of multiple screens to en sure that subjects complied with the sequence of treatments and measurements. The instruments were arranged such that first few pages measured subjects expertise a nd involvement in the experimental task, the following set of pages exposed subjects t o the first Web page and measured their corresponding perceptions, and the last set of page s exposed them to the second Web page, measured their corresponding perceptions, and required them to complete the experimental task. The expertand community-governed pages used for t he experiment were created using an open-source content management software, w hich was installed on the desktop
151 computer of the researcher. After the pages were c reated, they were converted to image files (to jpeg format) and were uploaded to a file server on the Web. The links of these image files were provided in the appropriate sectio ns of the measurement instrument for subjects to click and open. Although it was possib le to deploy the content management software on the Web and provide the links of these live pages in the measurement instrument, the pages were provided as images to su bject. This was because if the content management software was accessible on the W eb, subjects could search and find the Web pages assigned to other groups, jeopardizin g the internal validity of the experiment. Procedure Subjects participating in the experiment were rando mly assigned to one of the experimental groups, and were sent e-mails to infor m them of the corresponding link for their assigned group. Clicking the link directed s ubjects to the first page of the measurement instrument, which presented the instruc tions for the experiment. The instructions stated that subjects were planning a v isit to Cambodia for leisure, and were trying to gather travel related information about C ambodia on the Web. Subjects were told that their efforts to find information resulte d in two Web pages, which would be presented in the following pages. The instructions asked subjects to examine these two Web pages carefully, and answer the upcoming questi ons in the questionnaire. Subjects were also instructed that they would be required to create their travel plan based on the information provided on these Web pages at the end of the questionnaire. The set of instructions used in the experiment are provided in Appendix B.
152 Before subjects were exposed to the two Web pages, their expertise and involvement about Cambodia (i.e., their elaboration likelihood) were measured. Following this, subjects were presented with the li nk of the first Web page, which was configured to be opened in a new browser window or tab. Subjects were advised not to close that window or tab since they would need to r efer back to it to answer the upcoming questions. The related instructions provided to su bjects are presented in Appendix C. Upon clicking the link on the page, subjects were e xposed to the first Web page, which could be one of the pages presented in Appendix A d epending on the group they were assigned to. The next page of the questionnaire involved compreh ension questions to test whether subjects read and understood the Web page. There were a total of 11 comprehension questions per page, six of which were related to the governance mechanism used for that page, and five of which con cerned the topics discussed on the page. Sample comprehension questions used for the expert-governed page are presented in Appendix D. The bodies of all four Web pages included the same five topics about Cambodia: visa requirements, how to get there, where to stay, what to see, and where to exchange money. Appendix A presents all the Web pages used in the experiment. The experimental task was specifically chosen for creat ing a travel plan for a foreign country, because it is very common for individuals, even for students, to gather information from knowledge repositories before visiting a foreign co untry. The choice of country was motivated by the fact that Cambodia is not a very p opular destination for tourism compared to other European or Asian countries. If subjects had less expertise or
153 involvement about the task, they would rely on the peripheral route in making decisions, which would help test the effects of the credibilit y of governance mechanism construct. The four pages used for this experiment included si milar information about Cambodia to ensure that knowledge quality remained the same. However, pages made different suggestions on all five topics. For exam ple, in a given group, the first page suggested that tourists should visit the archeologi cal place Angkor Wat instead of Preah Khan (because Preah Khan is a smaller temple than Angkor Wat ), but the second page suggested that tourists should visit Bayon instead of Preah Khan (because Preah Khan is a smaller temple than Bayon ). While the first page did not discuss Bayon the second page did not discuss Angkor Wat The information on the Web pages were intentiona lly incomplete (rather than conflicting) because: (1) i ncomplete information is prevalent in many knowledge repositories since it may not be pos sible for knowledge contributors to cover all aspects of a phenomenon in detail; (2) co nflicting information could confuse subjects, and thus confound the results. The natur e of information provided on the pages can be found in Appendix A. In the last phase of the experiment, subjects were asked to complete the experimental task, which involved creating their tr avel plan for Cambodia based on the five topics discussed on the pages. For each topic subjects could choose the suggestion made by either of the pages they were given. The i nstructions for completing the experimental task and the related questions provide d to subjects are presented in Appendix E.
154 Operationalization of Constructs The constructs of interest in study are: elaboratio n likelihood, source credibility, knowledge quality, credibility of governance mechan ism, intention, and knowledge use. All constructs were measured using pre-validated it ems from prior research, but were reworded where necessary to fit the context of this s tudy. The measurement items for all constructs are presented in Table 18. Elaboration likelihood was measured using two separ ate constructs: subjects expertise and involvement in Cambodian tourism. Bo th constructs consisted of three Likert-scaled items adapted from Sussman and Siegal (2003) and Zaichkowsky (1985). Expertise concerned subjects level of knowledge ab out Cambodia and Cambodian tourism; whereas involvement concerned the degree t o which individuals were concerned about information on Cambodia or perceived it as im portant or relevant. It is important to note that the experimental task was designed to minimize subjects expertise and involvement in the subject matter. Therefore, thes e constructs were measured as control variables for the purposes of this study. Source credibility was measured using four Likert-s caled items adapted from Sussman and Siegal (2003). The items concerned the degree of knowledge, expertise, trustworthiness, and reliability of the individual who authored the information on the Web pages. Source credibility was not manipulated in the experiment and kept constant across all treatments. Therefore, source credibili ty was measured as a control variable.
155 User Expertise: (adapted from Sussman and Siegal 20 03) EXP1 How informed are you about Cambodia? Novice 1 2 3 4 5 6 7 Expert EXP2 To what extent are you an expert on Cambodia? Not at all 1 2 3 4 5 6 7 To a great ex tent EXP3 How informed are you about Cambodian tourism? Novice 1 2 3 4 5 6 7 Expert User Involvement: (adapted from Zaichkowsky 1985) Information about Cambodia is ____________ for me. INV1 Not important 1 2 3 4 5 6 7 Important INV2 Of no concern 1 2 3 4 5 6 7 Of concern INV3 Irrelevant 1 2 3 4 5 6 7 Relevant Source credibility: (adapted from Sussman and Siega l 2003) The person, who made the submission, is _______ abo ut Cambodia. SRC_CRED1 Knowledgeable 1 2 3 4 5 6 7 SRC_CRED2 Expert 1 2 3 4 5 6 7 SRC_CRED3 Trustworthy 1 2 3 4 5 6 7 SRC_CRED4 Reliable 1 2 3 4 5 6 7 Knowledge Quality: (adapted from Bhattacherjee and Sanford 2006) The information on the Web page about Cambodia is _____________. Strongly Neutral Strongly disagree agre e QUAL1 Informative 1 2 3 4 5 6 7 QUAL2 Helpful 1 2 3 4 5 6 7 QUAL3 Valuable 1 2 3 4 5 6 7 QUAL4 Persuasive 1 2 3 4 5 6 7 Credibility of governance mechanism: (adapted from Sussman and Siegal 2003) This Web page about Cambodia is ________. GOV_CRED1 Trustworthy 1 2 3 4 5 6 7 GOV_CRED2 Reliable 1 2 3 4 5 6 7 The (expert/users of the site), who (examined/exami ned and edited the submission), (is/are) __________ about Cambodia GOV_CRED3 Knowledgeable 1 2 3 4 5 6 7 GOV_CRED4 Expert 1 2 3 4 5 6 7 GOV_CRED5 Trustworthy 1 2 3 4 5 6 7 GOV_CRED6 Reliable 1 2 3 4 5 6 7 Intention: (adapted from Ajzen 2002) If I were going to Cambodia, I would ______ to use the information on the Web page. INT1 Intend 1 2 3 4 5 6 7 INT2 Try 1 2 3 4 5 6 7 INT3 Plan 1 2 3 4 5 6 7 Knowledge Use: (number of suggestions used from a s ingle Web page) / 5 Table 18. Measurement Items
156 Knowledge quality was measured using four Likert-sc aled items adapted from Bhattacherjee and Sanford (2006). The items tapped into the informativeness, helpfulness, value, and persuasiveness of the Web p age presented to subjects. These items were preferred over the three items (that con cerned completeness, accuracy, and consistency of knowledge) used in mainstream ELM re search, because subjects were not experts about the experimental task, and were unabl e to make such judgments about the Web pages used in the experiment. Credibility of governance mechanism was measured us ing six Likert-scaled items adapted from Sussman and Siegal (2003). Preliminar y interviews with several knowledge workers revealed that credibility of a go vernance mechanism consisted of two sub-dimensions: (1) the credibility of the individuals who performed the governance function; (2) the credibility of the page as a result of the governance process. For example, knowledge workers may perceive expert-gove rnance credible if the experts who perform the governance functions are credible, or i f the governance functions produce a credible knowledge asset. These two dimensions can vary independent of each other as a specific instance of governance mechanism can emplo y a credible set of governors, but produce a less credible knowledge asset due to poor ly executed governance functions. In order not to jeopardize the internal validity of th is study, these two dimensions were manipulated simultaneously in the same direction fo r creating the high and the low credibility conditions. Therefore, four items were used to measure the degree of knowledge, expertise, trustworthiness, and reliabil ity of the individuals who performed the governance function, and two items were used to measure the trustworthiness and reliability of the Web page.
157 Intention to use knowledge was adapted from Ajzen ( 2002), and measured using three Likert-scaled items. Subjects were asked whe ther they would intend, try, and plan to use the information provided on the Web pages if they were going to go to Cambodia. Knowledge use the dependent variable was measur ed as the percentage of the suggestions used from a single page. Since subjec ts were provided with five different topics, they could choose one of the suggestions fr om one of the pages per topic. This created two measures of knowledge use for each subj ect, one for the first Web page, the other for the second Web page. For example, if a s ubject used all five suggestions offered by the expert-governed page (but none offer ed by the community-governed page) in creating his/her itinerary, the subjects knowle dge use measures for the expertand the community-governed pages would be 100% and 0% respe ctively. Findings Pilot experiment A pilot experiment was conducted in late 2009 with 46 undergraduate students enrolled in a MIS course. Participation was volunt ary and students received extra credit for taking part in the study. The goal of the pilo t experiment was ensure that the credibility of governance mechanism could be manipu lated successfully. The pilot experiment was conducted in the same way the actual experiment was conducted with the exception that subjects received only one Web page (as opposed to two as in the actual experiment). There were a tot al of four groups, each receiving one credibility condition (high or low) for one of the governance mechanisms. For example, first group received an expert-governed page with h igh credibility condition, the second group received the same page with low credibility c ondition, and so on.
158 The subjects provided answers to the same measureme nt instrument developed for the actual experiment (except some items used a 6-p oint scale instead of the 7-point scale used in the actual experiment). The manipulation c heck, using one-way ANOVA, showed that the mean credibility scores of the gove rnance mechanisms across the four groups were significantly different from each other (Global-F=22.13; p<0.0001), as presented in Table 19. Group Mean Standard Deviation Expert-governance Low credibility (EG-L) 1.82 0.8 8 Expert-governance High Credibility (EG-H) 4.95 0. 80 Community-governance Low Credibility (CG-L) 2.96 1.35 Community-governance High Credibility (CG-H) 4.25 0.75 Table 19. Pilot experiment descriptive statistics The pair-wise comparisons between the groups reveal ed that both expertgovernance and community-governance were successful ly manipulated with statistical significance. In expert-governance, the mean credi bility scores of high and low conditions were 4.95 and 1.82 (out of 6) respective ly, and the difference was statistically significant (p<0.0001). Similarly, in community-go vernance, the mean credibility scores of high and low conditions were 4.25 and 2.96 (out of 6) respectively, and the difference was statistically significant (p<0.015). The mean credibility scores of the four groups are plotted in Figure 11.
159 Figure 11. Mean credibility scores of governance me chanisms in the pilot experiment It is worth mentioning that the findings of the pil ot experiment are based on a low sample size with the credibility construct violatin g the normality assumption. Although ANOVA is considered robust with respect to normalit y (O'Brien, 1979), the data was reexamined using the Kruskal-Wallis test, which is a non-parametric test for non-normal data. The findings still suggested that the differ ences in means were significant (Chisquare=26.58; p<0.0001). Two important insights were gained from the pilot e xperiment: (1) the sevenpoint scale would have been a more appropriate meas urement scale instead of the sixpoint scale, as subjects could not select neutral for non-manipulated constructs such as source credibility; (2) the manipulation needed ref inement to further increase the differences in means of the high and the low credib ility conditions. Experiment The actual experiment was conducted in January and February of 2010. In order to determine if the experiment needed to be complet ed in a controlled laboratory or not, two initial sessions were held. In the first sessi on, 49 subjects participated in the 0 1 2 3 4 5 6 EG-LEG-HCG-LCG-H Credibility rating Governance Mechanisms
160 experiment in a computer laboratory with the existe nce of the researcher. In the second session, 38 students participated in the experiment completely online at their own convenience. The results of these two sessions wer e the same. The mean comprehension scores of subjects were 95% in the laboratory sessi on and 94% in the online session (p=0.45). Therefore, the experiment proceeded with a third and completely online session to increase participation. The experiment was run for a total of two weeks, for which 370 responses were collected from a total of 555 students. Combining all three sessions, the study collected responses from 457 su bjects out of a possible 648. The mean comprehension score of subjects was 95% wi th a standard deviation of 9.2. Using the three standard deviations of the me an as a cut-off line, a score of 67% was determined as the borderline for the validity of a response. Accordingly, nine responses (out of 457) were flagged as invalid since the comp rehension scores of those subjects were below 67%. Furthermore, five subjects rated t heir expertise as being higher than four (on a seven-point scale), which posed a threat for the activation of the central route instead of the peripheral route in answering questi ons. Dropping these subjects further from the data set brought the usable number of resp onses to 443 for data analysis. Outlier analysis Prior to analyzing the data, an outlier analysis wa s conducted at both univariate and multivariate levels. For univariate outliers, each measurement item was examined separately in each group. Accordingly, the mean an d standard deviation of an item were calculated, and responses that were outside three s tandard deviations of the mean were flagged as outliers. On the other hand, multivaria te outliers were examined in the multi-
161 dimensional space resulting from the joint combinat ion of all items in a single group. Since the unidimensional approach does not apply to a multi-dimensional space, the Mahalanobis distance was used to identify outliers (Penny, 1996). This statistic merely represents the distance of a single observation fro m the center of the cluster formed by the rest of the observations. The larger the stati stic, the more likely the observation is an outlier; because a large distance indicates that th e observation is farther away from the rest of the observations. In order to calculate th is statistic, all measurement items in a single group were regressed on the knowledge use va riable measured for that group. At the end of the outlier analysis, 20 observations were flagged as outliers. Sixteen of these were at the univariate level, two were at the multivariate level, and two were at both the univariate and the multivariate le vels. A closer examination of these observations revealed that subjects gave random ans wers to questions (such as all seven or all one) although they scored well on the compre hension questions. Therefore, these observations were dropped from the data set, bringi ng the total number of usable observations to 423. The distribution of these res ponses across the experimental groups is presented in Table 20 with the mean comprehensio n score in each group. The descriptive statistics of each measurement item are presented in Table 21. Group No. of subjects Mean comprehension score Standard deviation 1 96 97% 0.06 2 103 95% 0.06 3 108 96% 0.06 4 116 95% 0.07 ALL 423 96% 0.06 Table 20. Distribution of subjects within groups
162 Expert-Governance Community-Governance Mean Std.Dev. Skewness Kurtosis Mean Std.Dev. Skewn ess Kurtosis EXP1 1.46 0.86 1.96 2.93 EXP2 1.17 0.56 3.59 13.05 EXP3 1.20 0.54 2.98 9.10 INV1 2.80 1.44 0.04 -1.04 INV2 2.85 1.42 0.01 -1.06 INV3 2.84 1.42 -0.01 -1.09 GOV_CRED_1 4.12 1.69 -0.24 -0.92 4.12 1.62 -0.45 -0 .81 GOV_CRED_2 4.06 1.70 -0.16 -1.01 4.10 1.66 -0.34 -0 .89 GOV_CRED_3 4.13 2.24 -0.22 -1.49 4.26 1.89 -0.46 -1 .09 GOV_CRED_4 3.72 2.21 0.06 -1.50 3.46 1.75 -0.02 -1. 23 GOV_CRED_5 3.79 2.05 -0.07 -1.40 3.98 1.65 -0.40 -0 .80 GOV_CRED_6 3.78 2.12 -0.04 -1.47 3.96 1.71 -0.31 -0 .96 SRC_CRED_1 4.23 1.36 -0.52 -0.31 4.58 1.31 -0.91 0. 28 SRC_CRED_2 3.02 1.39 0.40 -0.43 3.19 1.44 0.15 -0.8 8 SRC_CRED_3 4.04 1.27 -0.35 -0.10 4.25 1.23 -0.55 0. 38 SRC_CRED_4 3.96 1.31 -0.31 -0.27 4.19 1.28 -0.50 -0 .01 QUAL_1 4.98 1.27 -1.11 1.55 5.07 1.31 -1.31 1.77 QUAL_2 4.89 1.30 -0.93 0.92 5.03 1.33 -1.27 1.51 QUAL_3 4.49 1.48 -0.50 -0.29 4.56 1.47 -0.67 -0.04 QUAL_4 4.05 1.61 -0.21 -0.84 4.22 1.59 -0.41 -0.69 INT_1 3.91 1.78 -0.10 -1.10 4.08 1.68 -0.30 -0.91 INT_2 4.21 1.75 -0.34 -0.95 4.31 1.69 -0.45 -0.73 INT_3 3.75 1.83 0.00 -1.17 3.84 1.75 -0.08 -0.98 Legend: EXP : Expertise; INV : Involvement; GOV CRED : Credibility of governance mechanism; SRC CRED : Source credibility, QUAL : Quality; INT : Intention Table 21. Descriptive statistics of measurement ite ms Manipulation check In order to see if high and low credibility conditi ons were successfully created, two different manipulation checks were conducted, o ne for expert-governance, and another for community-governance, using the respons es provided to the questions concerning credibility of governance mechanisms (pl ease see Table 18 for the related
163 questions). The manipulation check for expert-gove rnance revealed that subjects were successfully assigned to high and low credibility c onditions, as the mean credibility score of the high condition was higher than that of the l ow condition with statistical significance (5.53 versus 2.52 respectively; p<0.00 01). The manipulation check for community-governance yielded similar results (5.05 for the high condition versus 2.98 for the low condition; p<0.0001), suggesting that s ubjects were assigned to high and low credibility conditions successfully. These finding s are summarized in Table 22. Governance mechanism Number of subjects Mean credibility score (Std.dev) Significance EG-H 199 5.53 (0.94) p<0.0001 EG-L 224 2.52 (1.18) CG-H 204 5.05 (0.93) p<0.0001 CG-L 219 2.98 (1.32) Legend: EG-H : Expert-governance with high credibility; EG-L : Expertgovernance with low credibility; CG-H : Community-governance with high credibility; CG-L : Community-governance with low credibility; Std.dev: Standard deviation Table 22. Results of the manipulation check Order effects The experimental design is susceptible to order eff ects, because the measurements for expertand community-governed pages were taken sequentially rather than simultaneously. The use of the counterbalanced des ign allows checking for the order effects and their potential influence on the findin gs of this study. Before analyzing the order effects, it is important to clarify a misconc eption about counterbalanced designs. Researchers tend to think that counterbalancing co ntrols the measured variable(s) since combining the responses obtained from a certain tre atment sequence with the responses
164 obtained from the reverse sequence cancels the ef fects of the order of treatments. However, it has been suggested that counterbalancin g can control a variable only if there is no interaction between the counterbalanced variable and the order of treatments (Keppel, 1991; Reese, 1997; Winer et al., 1991). I f there is an interaction effect, it means that the first treatment affects the second treatme nt, and responses given to the second treatment are plagued with adaptation, fatigue, or other types of carry-over problems. This can be explained using Figure 12. Imagine a c ounterbalanced design for two treatments. The first group receives treatment 1 f irst and treatment 2 second, while the second group receives treatment 2 first and treatme nt 1 second. The mean values of a specific variable for the two treatments across the two groups are plotted in the left panel of Figure 12. If the variable is not influenced by the order of treatments, the slopes of the measurements are the same, and counterbalanced desi gn ensures that the differences in the means of the two treatments are equal across th e two groups. If the differences in means vary across the two groups, as shown in the r ight panel of Figure 12, the order of treatments influences the variable. This changes t he slopes, and suggests that the effects of the first treatment are transferred over to the effects of the second treatment. Therefore, counterbalanced designs help researchers check for the potentially confounding effects of order of treatments using th e interaction term, and determine whether the findings are meaningful and interpretab le. It has also been suggested that the main effects of the treatments (in addition to the interaction term) should be checked for correct interpretation of findings. For example in the left panel of Figure 12, A and C (and B and D) should not be statistically different from each other. However, researchers argue that when there are only two treatments (as i n this study), this restriction can be
165 relaxed and researchers can only check for interact ion effects (Keppel, 1991; Reese, 1997). Figure 12. Effects of counterbalancing on measureme nt In light of the above suggestions, repeated measure s ANOVA was conducted for each variable using the order of treatments as a be tween-subject factor for each distinct group. The findings, as presented in Table 23, sho wed that some of the variables interacted with the order of treatments suggesting a transfer of effects from the first treatment to the second. For example, the interact ion term of source credibility was consistently significant across all groups. Simila rly, knowledge use (the dependent variable of this study) had a significant interacti on with the order of treatments for the two groups in which both treatments were set to eit her high or low credibility conditions simultaneously. Further, the credibility of govern ance mechanism, intention, and quality constructs in the last two groups had significant i nteraction terms signaling a transfer of effects. These findings were both baffling and int eresting, because they provided insights Mean values of the variable Mean values of the variable Treatment Treatment
166 for understanding the complexity of knowledge use f rom electronic repositories. A more detailed interpretation of these findings is provid ed in the Post-hoc Analysis section of this essay. Expert-governed page Community-governed page Significance of interaction effect Group Variables Mean (first in sequence) Mean (second in sequence) Mean (first in sequence) Mean (second in sequence) EG-H vs. CG-H Quality 4.93 5.14 5.38 5.14 ns. Gov.cred. 5.43 5.42 4.85 5.03 ns. Src.cred. 4.14 3.56 4.60 4.33 ** Intention 4.51 4.87 4.79 4.81 ns. Know.use 0.33 0.53 0.47 0.67 EG-H vs. CG-L Quality 5.11 5.63 4.54 3.53 ns. Gov.cred. 5.30 5.90 3.34 2.48 ns. Src.cred. 4.01 3.86 4.00 3.08 ** Intention 4.38 5.45 3.61 2.68 ns. Know.use 0.80 0.81 0.19 0.20 ns. EG-L vs. CG-H Quality 4.38 3.68 5.28 5.49 ns. Gov.cred. 2.32 2.35 4.88 5.39 Src.cred. 3.88 3.48 4.58 4.35 Intention 3.34 2.61 4.75 5.13 ns. Know.use 0.10 0.11 0.89 0.90 ns. EG-L vs. CG-L Quality 4.23 3.93 4.56 3.79 Gov.cred. 2.32 2.98 3.45 2.50 ns. Src.cred. 4.07 3.61 3.90 3.56 ** Intention 3.27 3.40 3.80 3.00 Know.use 0.53 0.29 0.71 0.47 ** Legend : EG : expert-governance; CG : community-governance, H : high credibility condition; L : low credibility condition (*): p<0.05; (**): p<0.0001; ns.: non-significant Table 23. Order effects Although the above findings suggest that the order of treatments interacted with certain variables, these variables were not dropped from the analysis for a couple of
167 reasons. First, the order of treatments reflects w hat really transpires during the actual use of knowledge from electronic repositories. In actu al knowledge use situations, individuals make decisions after they retrieve know ledge from repositories sequentially and in random order without a predefined sequence. Therefore, the experimental design can be considered a proxy of actual knowledge use s ituations, and counterbalancing helps us understand how individuals react to different ty pes of knowledge sources if these sources are encountered in a certain sequence. Sec ond, the order effects were taken into account by analyzing the data for each treatment se quence instead of pooling the data of the counterbalanced groups. This ensured that orde r effects were contained during hypotheses testing, and did not plague the results. Therefore, none of the observations or variables was dropped from the analysis. Scale validity Confirmatory factor analysis (CFA) was used to test the reliability and the construct validity of the measurement items used in this study. CFA was preferred over the exploratory factor analysis (EFA), because late nt constructs were informed by a priori theory and the measurement instrument used pre-val idated items (Bagozzi and Phillips, 1982). Therefore, all items were modeled as indicators of their corresponding latent constructs, and all constructs were allowed to covary among themselves. The scale validity of measurement items used in the experiment was assessed using convergent and discriminant validity. Conver gent validity was determined using three criteria as suggested by Fornell and Larcker (1981): (1) all factor loading should be significant and higher than 0.7; (2) composite reli ability (c) of each construct should
168 exceed 0.8, and (3) the average variance extracted (AVE) for each construct should exceed 0.5 (or the square root of AVE for each cons truct should exceed 0.71). Discriminant validity was also assessed in light of Fornell and Larckers (1981) suggestion, which stated that the square root of AV E for each construct should exceed the correlation of that construct with other constructs Expert-governed Web page (first in sequence) Expert-governed Web page (second in sequence) Mean Std.Dev. Loading (*) Mean Std.Dev. Loading (*) EXP1 1.50 0.88 0.66 EXP2 1.20 0.58 0.66 EXP3 1.22 0.54 0.95 INV1 2.82 1.45 0.94 INV2 2.94 1.45 0.97 INV3 2.89 1.44 0.98 GOV_CRED1 4.01 1.61 0.75 4.21 1.75 0.83 GOV_CRED2 3.93 1.63 0.66 4.18 1.74 0.84 GOV_CRED3 3.90 2.29 0.89 4.35 2.17 0.95 GOV_CRED4 3.42 2.20 0.93 4.00 2.18 0.93 GOV_CRED5 3.53 2.06 0.91 4.03 2.02 0.98 GOV_CRED6 3.51 2.14 0.93 4.04 2.08 0.99 SRC_CRED1 4.45 1.36 0.83 4.03 1.34 0.75 SRC_CRED2 3.12 1.46 0.64 2.93 1.32 0.59 SRC_CRED3 4.28 1.21 0.96 3.83 1.28 0.96 SRC_CRED4 4.20 1.24 0.87 3.74 1.33 0.92 QUAL1 5.04 1.24 0.94 4.93 1.30 0.91 QUAL2 5.01 1.22 0.80 4.78 1.36 0.94 QUAL3 4.54 1.42 0.88 4.45 1.53 0.95 QUAL4 3.94 1.63 0.88 4.16 1.58 0.80 INT1 3.72 1.77 0.93 4.07 1.78 0.98 INT2 4.20 1.74 0.96 4.23 1.76 0.95 INT3 3.57 1.87 0.92 3.93 1.78 0.95 Legend : EXP : Expertise; INV : Involvement; GOV CRED : Credibility of governance mechanism; SRC CRED : Source credibility, QUAL : Quality; INT : Intention (*): Significant at p<0.0001 Table 24. Factor loadings of items used for the exp ert-governed page
169 Scale validity of the measurement items used for th e two pages was assessed separately and for both treatment sequences (due to order effects). The factor loadings of the items used for the expert-governed page are pre sented in Table 24. The left panel of the table shows the loadings when the page was give n first, and the right panel of the table shows the loadings when the page was given se cond. The item loadings of the community-governed page are presented in the left a nd the right panels of Table 25. Community-governed Web page (first in sequence) Community-governed Web page (second in sequence) Mean Std.Dev. Loading (*) Mean Std.Dev. Loading (*) EXP1 1.42 0.85 0.64 EXP2 1.15 0.54 0.68 EXP3 1.19 0.54 0.83 INV1 2.79 1.43 0.81 INV2 2.77 1.40 0.92 INV3 2.79 1.39 0.95 GOV_CRED1 4.22 1.44 0.72 4.01 1.80 0.87 GOV_CRED2 4.19 1.48 0.55 4.01 1.84 0.88 GOV_CRED3 4.34 1.75 0.93 4.16 2.04 0.92 GOV_CRED4 3.44 1.60 0.95 3.49 1.91 0.91 GOV_CRED5 4.10 1.46 0.94 3.86 1.84 0.97 GOV_CRED6 4.10 1.51 0.97 3.81 1.90 0.97 SRC_CRED1 4.85 1.24 0.83 4.29 1.33 0.78 SRC_CRED2 3.33 1.46 0.57 3.03 1.41 0.66 SRC_CRED3 4.40 1.16 0.94 4.09 1.29 0.92 SRC_CRED4 4.38 1.17 0.88 3.99 1.36 0.97 QUAL1 5.28 1.18 0.93 4.84 1.41 0.91 QUAL2 5.29 1.15 0.77 4.74 1.44 0.95 QUAL3 4.74 1.35 0.76 4.37 1.57 0.93 QUAL4 4.31 1.50 0.76 4.13 1.69 0.85 INT1 4.16 1.59 0.94 3.99 1.77 0.96 INT2 4.46 1.56 0.98 4.14 1.80 0.92 INT3 3.95 1.67 0.93 3.71 1.83 0.95 Legend : EXP : Expertise; INV : Involvement; GOV CRED : Credibility of governance mechanism; SRC CRED : Source credibility, QUAL : Quality; INT : Intention (*): Significant at p<0.0001 Table 25. Factor loadings of items used for the com munity-governed page
170 As seen in both tables, all item loadings were sign ificant and met the minimum loading criterion except a few. Items that had poo r loading were the same for both expertand community-governed pages, and included the second item of the credibility of governance mechanism construct (when pages were given first to the subjects), the second item of the source credibility construct (fo r both sequences), and the first two items of the expertise construct. The second condition of convergent validity was ass essed by checking the composite reliability of each construct for both ex pert and community-governed pages. The composite reliability score of each construct a nd the correlation of that construct with other constructs are presented in Table 26 for the expert-governed page and Table 27 for the community-governed page. The left panel s of both tables show the results when the pages were given first to the subjects, an d the right panel shows the results when they were given second. Expert-governed Web page (first in sequence) Expert-governed Web page (second in sequence) c 1 2 3 4 5 6 c 3 4 5 6 1 EXP 0.86 0.86 2 INV 0.96 0.24 0.94 3 GOV_CRED 0.95 0.10 0.04 0.87 0.97 0.92 4 SRC_CRED 0.88 0.10 0.08 0.12 0.82 0.88 0.28 0.82 5 QUAL 0.89 -0.03 0.03 0.40 0.54 0.84 0.94 0.72 0.54 0.90 6 INT 0.94 0.07 0.02 0.41 0.60 0.67 0.92 0.97 0.75 0.45 0.80 0.96 Diagonal elements represent the square root of AVE for each construct c = Composite reliability Table 26. Composite reliability, AVE, and correlati ons for the expert-governed page As shown in both tables, all composite reliability scores were higher than 0.8 for both sequences of pages (with the experience constr uct having the lowest score of 0.82
171 for the community-governed page when it was first i n sequence). The third, and the final, condition of convergent validity was assessed by ch ecking the AVE value of each construct. All AVE values, as the diagonal element s in Table 26 and Table 27, were higher than 0.71 (the lowest being the credibility of governance and source credibility constructs for the community-governed page with an AVE value of 0.81). Community-governed Web page (first in sequence) Community-governed Web page (second in sequence) c 1 2 3 4 5 6 c 3 4 5 6 1 EXP 0.82 0.87 2 INV 0.96 0.25 0.95 3 GOV_CRED 0.92 0.10 0.10 0.81 0.97 0.92 4 SRC_CRED 0.86 0.08 0.02 0.63 0.81 0.90 0.71 0.84 5 QUAL 0.89 -0.07 0.01 0.53 0.69 0.83 0.95 0.77 0.76 0.91 6 INT 0.94 0.05 0.02 0.52 0.64 0.65 0.92 0.96 0.82 0.71 0.80 0.94 Diagonal elements represent the square root of AVE for each construct c = Composite reliability Table 27. Composite reliability, AVE, and correlati ons for the community-governed page Finally, discriminant validity was assessed by co mparing the square root of AVE for each construct to the correlation of that const ruct with other constructs. For the expert-governed page, the lowest square root of AVE which was 0.82 for source credibility, was higher than the highest correlatio n among factors, which was 0.80 between intention and quality constructs. Similarl y, for the community-governed page, the lowest square root of AVE was 0.81 for the cred ibility of governance construct (as well as source credibility), which was larger than the highest correlation of 0.80 between intention and quality. These findings suggested th at discriminant validity criterion was also satisfied.
172 As a result of scale validation, the two items of t he expertise construct, the second item of the credibility of governance mechanism con struct, and the second item of the source credibility construct were excluded from fur ther analysis since they violated the convergent validity criterion. Hypotheses testing The next step of the analysis was to test the hypot heses posited earlier. Since the preliminary analysis revealed order effects, hypoth eses were tested for treatment sequence separately. The analysis was conducted us ing partial least squares (PLS) provided by the SmartPLS software package (Ringle e t al., 2005). The selection of PLS over covariance-based structural equation modeling was motivated by two reasons: (1) PLS can handle the moderating effects of expertise and involvement (if there are any) better than covariance-based structural equation mo deling; (2) PLS is not sensitive to the distributional assumptions commonly made in covaria nce-based structural equation modeling. Before proceeding to results, three non-manipulated constructs deserve further attention: source credibility, expertise, and invol vement. The experiment was designed such that none of these constructs should have show n any variation between or within subjects. There were two major reasons for this: ( 1) the information about the source (i.e., the contributor) of each Web page was kept t he same throughout the experiment to eliminate any confounding effects of source credibi lity on the dependent variable; and (2) the task was chosen specifically to minimize subjec ts expertise and involvement in the subject matter to invoke their peripheral route rat her than their central route.
173 The preliminary analysis showed that subjects perc eptions of source credibility differed within and between groups. Therefore, sou rce credibility was included into the analysis as a control variable. However, analyzing the effects of expertise and involvement showed that all interaction effects ass ociated with these constructs were non-significant for both the expertand the commun ity-governed page. This was because neither expertise nor involvement showed any variat ion within or between groups. The results of the interaction effects are presented in Appendix F. In order to ensure that the interaction effects were insignificant, the effect sizes ( f ) of the interaction effects on intention were computed using Cohen and Cohens (19 83) formula ( f = [R2 interaction effects model R2 main effects model] / [R2 interaction effects model]). The corresponding improvements in the R2 value of the intention construct with and without the interaction effects, and the resulting effects sizes are presented in Table 28. R2 of intention without expertise and involvement R2 of intention with expertise and involvement Effect size ( f ) Expert-governed page First in sequence 0.50 0.53 0.06 Second in sequence 0.75 0.75 0.03 Community-governed page First in sequence 0.76 0.77 0.01 Second in sequence 0.52 0.58 0.10 Table 28. Comparison of interaction models with mai n effects models As seen in the table, some of the effects were smal l to moderate, suggestion that they be included into the model (Wynne et al., 2003 ). However, the interaction effects
174 were still dropped from data analysis for the sake of parsimony, since their path coefficients were consistently non-significant. The first phase of model testing concerned the rela tionships proposed for the expert-governed page. Due to the existence of orde r effects, hypotheses were tested for both page sequences. The findings are presented in Figure 13. In the figure, the values without parentheses are for the case when subjects were exposed to the expert-governed page first (hereafter referred to as EG1), while th e values with the parentheses are for the case when subjects were exposed to the expert-gover ned page second (hereafter referred to as EG2). As shown in the figure, all hypotheses were supported for both cases. Notes: 1) (*): p<0.05 2) Values without parentheses: subjects were given the expert-governed page first (EG1); Values with parentheses: subjects were given t he expert-governed page second (EG2). Figure 13. Parameter estimates of expert-governance model
175 In line with prior research, intention to use knowl edge had a positive and significant effect (EG1 = 0.23 and EG2 = 0.59; p<0.05) on the actual use of knowledge supporting H1a. As hypothesized in this essay, cre dibility of expert-governance positively affected both intentions to use knowledg e (EG1 = 0.21 and EG2 = 0.41; p<0.05) supporting H2a, and perceptions of knowledg e quality (EG1 = 0.40 and EG2 = 0.65; p<0.05) supporting H3a. All non-hypothesized relationships were in line wit h expectations and with prior research. The effect of knowledge quality on inten tion was positive and significant (EG1= 0.55 and EG2 = 0.47; p<0.05). Further, source credibility, as t he control variable, had a positive and significant effect on quality (EG1 = 0.43 and EG2 = 0.34; p<0.05), and a positive but non-significant effect on intention (EG1 = 0.05, p=0.48; EG2 = 0.06, p=0.15). The analysis of the community-governed page also yi elded similar results. The corresponding findings are presented in Figure 14, in which values without parentheses are for the case when subjects were exposed to the community-governed page second (hereafter referred to as CG2), while the values wi th the parentheses are for the case when subjects were exposed to the community-governe d page first (hereafter referred to as CG1). As expected, intention to use knowledge had a posit ive and significant effect on the actual use of knowledge (CG2 = 0.53 and CG1 = 0.26; p<0.05) supporting H1b. H2b was also supported since the effect of the credibil ity of community-governance on intention was positive and significant (CG2 = 0.40 and CG1=0.22; p<0.05). Finally, the
176 effect of the credibility of community-governance o n quality was positive and significant (CG2 = 0.49 and CG1 = 0.20; p<0.05), supporting H3b. The non-hypothesized relationships were in line wit h expectation, as quality had a positive and significant effect on intention (CG2 = 0.50 and CG1 = 0.45; p<0.05), and the control variable, source credibility, had a positiv e and significant effect on quality (CG2=0.40 and CG1 = 0.58; p<0.05), and a positive but non-significant effect on intention (CG2 = 0.03, p=0.37; CG1 = 0.15, p=0.11). Notes: 1) (*): p<0.05 2) Values without parentheses: subjects were given the community-governed page second (CG2); Values with parentheses: subjects were give th e community-governed page first (CG1). Figure 14. Parameter estimates of community-governa nce model Overall, the above findings support the notion that credibility of a governance mechanism is a salient peripheral route construct t hat influences individuals use of
177 knowledge from electronic repositories. It affects individuals perceptions of knowledge quality as well as their intentions, as hypothesize d in this study. Post-hoc analysis In addition to testing the hypotheses of this study a post-hoc analysis was conducted to gain more insights about individuals use of knowledge from repositories in the existence of governance mechanisms. Since, eac h subject was exposed to one expertgoverned and one community-governed Web page, parti cipants perceptions of the two pages were analyzed for each group. The analysis i nvolved the comparison of the means of the constructs relevant to the hypothesized rela tionships in the study. The experimental design prevented the possibility to us e ANOVA to make the comparisons, because the samples that were being compared were n ot independent. Therefore, repeated measures ANOVA was employed, which is the most commonly used technique to analyze the effects of interventions that involv e preand post-treatment measurements. The null hypothesis of repeated measures ANOVA mere ly states that there is no difference in the means of the first and the second measurement (H0: [first measurement mean second measurement mean] = 0). Since the preliminary analysis revealed that measur ement of variables were influenced by the order of the treatments, two sepa rate repeated measures ANOVA were conducted for each group, one for the case when sub jects were exposed to expertgoverned page first and community-governed page sec ond, and another for the case when the order was reversed. It is important to mention that no between-subject comparison was made, since such a comparison was non-interpret able. Among the within-subject comparisons, the below discussion focuses on only G roup 1 and Group 4. This is
178 because the findings in these groups were more inte resting as the pages used in these groups were set to the same credibility condition ( i.e., Group 1 received both pages with high credibility condition; and Group 4 received bo th pages with low credibility condition). On the other hand, subjects in Group 2 and Group 3 received one highcredibility and one low-credibility page, which led individuals to have more favorable perceptions for the high-credibility page regardles s of whether the page was governed by expertor community-governance. Therefore, the be low discussion involves the withinsubject comparisons for Group 1 and Group 4. The first analysis involved the credibility of gove rnance mechanism, for which the findings are plotted in Figure 15. The left pa nel of the figure shows the findings for Group 1 (which received both governance mechanisms with high credibility condition), while the right panel shows the findings for Group 4 (which received both governance mechanisms with low credibility condition). Both p anels show the mean scores of the credibility of governance mechanism construct for t he two sequences used in the experiment. For example, in the left panel, the da shed line represents the mean scores of credibility when subjects were given the expert-gov erned page first and the communitygoverned page second. On the other hand, the solid line represents the mean scores when subjects were given the community-governed page fir st and the expert-governed page second. The left panel of Figure 15 shows that when both go vernance mechanisms were set to high credibility condition, subjects perceiv ed expert-governance to be more credible than community-governance page regardless of the sequence of treatments. For example, when subjects were given the expert-govern ed page first and the community-
179 governed page second (i.e., the dashed line in the left panel of the figure), they rated the credibility score of expert-governance with a score of 5.43, and the credibility of community-governance with a score of 5.03. The sam e trend was observed for the reverse sequence, as subjects rated the credibility of community-governance with a lower score (4.85) than the credibility of expert-governa nce (5.42). (*): within-subject p<0.05 (ns): within-subject p-value is non-significant Figure 15. Repeated measures ANOVA for credibility of governance mechanism There are two possible explanations for this. Firs t, the manipulation might not have set the credibility of community-governance ap propriately to the high condition. In other words, it may be that cues used to create the high credibility condition for community-governance were inadequate or weaker comp ared to expert-governance. This is plausible, because the manipulation check that w as performed during the pilot experiment on independent groups of subjects signal ed a similar problem. The second explanation for this finding is that subjects appro ached more favorably toward expertgovernance than community-governance. The reason f or this could be that the 5.43 5.03 5.42 4.85 2 3 4 5 6 EG-HCG-H Credibility of governance mechanism EG-H first; CG-H second (*) CG-H first; EG-H second (*) 2.32 2.5 2.98 3.45 2 3 4 5 6 EG-LCG-L Credibility of governance mechanism EG-L first; CG-L second (ns) CG-L first; EG-L second (*)
180 involvement of a designated (and possibly an accred ited) expert in executing certain governance functions can supersede the involvement of community members in executing the same or similar governance functions, no matter how credible the community members can be. This is plausible, becau se individuals rely on accredited experts in most phases of their lives. For example we tend to follow the advice of physicians as opposed to individuals who experience certain ailments firsthand and offer working solutions, because physicians are accredite d to provide advice compared to others. Subjects of the experiment could be influe nced by the same phenomenon, as the expert in expert-governance was designated and accr edited by the provider of the repository, while the community-members were being vigilantes without a formal endorsement from the repository provider. This, in turn, led individuals to have more favorable perceptions toward expert-governance than community-governance regardless of the sequence of exposure to the governance mecha nisms. While the above explanation can be valid for credib le experts, the advantages of accreditation may disappear when experts lack credi bility. This is because community has an informational advantage over a single indivi dual even if neither the community members nor the expert are credible. The data prov ides support for this argument in the right panel of Figure 15, which shows the findings for the case when both governance mechanisms are set to low credibility condition. A s seen in the figure, subjects perceived the credibility of expert-governance to be lower th an community-governance regardless of the sequence of treatments. In this case, subje cts had a higher valuation of the collective wisdom and the effort of the community c ompared to the expert. In line with
181 the previous explanation, subjects might have discr edited the expert, but had more faith in community. However, it is important to note that this finding can also be an artifact of inadequate manipulation. As described earlier, the manipulation might not have set the low credibility condition of community-governance a ppropriately. If this is the case, subjects might have selected the neutral option f or their perceptions of the credibility of community-governance, indicating their indifference This could increase the credibility score and lead to the findings presented in the rig ht panel of the figure. For this reason, findings need to be interpreted cautiously. The second repeated measures ANOVA concerned knowle dge quality of the two Web pages provided to subjects. Knowledge quality was not manipulated in the experiment, as the contents of both pages looked an d read the same except the specific suggestions provided by each page. However, as see n in Figure 16, subjects had different quality perceptions for the pages. For example, wh en both mechanisms were set to high credibility condition, as seen in the left panel of the figure, subjects perceived the quality of the community-governed page as being higher than the quality of the expert-governed page. This is surprising, because quality percepti ons do not correlate with the credibility of governance mechanisms. For instance, this group of individuals (i.e., Group 1) perceived expert-governance as being more credible than community-governance, but they found the page provided by community-governanc e to be of higher quality. Unless this is a spurious finding, it suggests that subjec ts had a greater appreciation for the quality of community-governed knowledge assets than the quality of expert-governed knowledge assets. In other words, they may have be lieved that knowledge quality is
182 more likely to be increased by community members c ollective efforts than an experts individual efforts. (*): within-subject p<0.05 (ns): within-subject p-value is non-significant Figure 16. Repeated measures ANOVA for knowledge qu ality It is interesting to note that when both mechanisms were set to the low credibility condition (i.e., the right panel of the figure), su bjects were influenced by the order of treatments. They consistently perceived the second page as being lower in quality than the first page. This phenomenon is referred to as the recency effect (Asch, 1946), where individuals are more influenced by the last treatme nt they are given. Since both mechanisms were set to the low credibility conditio n, individuals may have undervalued the quality of the second treatment more since they had a more vivid memory of the credibility of the second treatment. The third repeated measures ANOVA involved the inte ntion construct. The findings, presented in Figure 17, suggest that subj ect intention to use knowledge from 4.93 5.14 5.14 5.38 2 3 4 5 6 EG-HCG-H Knowledge quality EG-H first; CG-H second (ns.) CG-H first; EG-H second (ns.) 4.23 3.79 3.93 4.56 2 3 4 5 6 EG-LCG-L Knowledge quality EG-L first; CG-L second (*) CG-L first; EG-L second (*)
183 the two pages was a function of the order of treatm ents (despite weak statistical support). It is worth mentioning that subjects in Group 2 and Group 3, whose results are not shown in the figure, had knowledge use intentions in the expected directions. They had higher levels of intention to use knowledge from the gover nance mechanism that was set to the high credibility condition. However, when both mec hanisms were set to the same credibility conditions, subjects were influenced by the recency effect. For example, when both mechanisms were set to the high credibility co nditions (i.e., the left panel of Figure 17), subjects had higher levels of intentions to us e knowledge from the second page. When both mechanisms were set to the low credibilit y condition (i.e., the right panel of Figure 17), subjects had higher levels of intention to use knowledge from the first page. This is interesting, because intention is not corre lated to the credibility of governance mechanisms or knowledge quality. (*): within-subject p<0.05 (ns): within-subject p-value is non-significant Figure 17. Repeated measures ANOVA for intention 4.51 4.81 4.87 4.79 2 3 4 5 6 EG-HCG-H Intention EG-H first; CG-H second (ns.) CG-H first; EG-H second (ns.) 3.27 3.00 3.40 3.80 2.00 3.00 4.00 5.00 6.00 EG-LCG-L Intention EG-L first; CG-L second (ns.) CG-L first; EG-L second (*)
184 The final repeated measures comparison involved sub jects use of knowledge from the two Web pages. The results for Group 2 an d Group 3 (i.e., when subjects were assigned to one high credibility and one low credib ility governance-mechanism) were in line with expectations such that subjects tended to use more knowledge from the Web page that was governed with a more credible mechani sms compared to a less credible one. On the other hand, when the governance mechan isms were set to the same credibility condition, subjects use of knowledge w as again influenced by recency effects. Accordingly, subjects used more knowledge from the second page when they were assigned to high credibility conditions for both go vernance mechanisms (the left panel of Figure 18). Similarly, they used less knowledge fr om the second page when they were assigned to the low credibility condition for both mechanisms (the right panel of Figure 18). It is also important to note that these findi ngs are consistent with subjects intentions to use knowledge as discussed in the previous parag raph. (*): within-subject p<0.05 (ns): within-subject p-value is non-significant Figure 18. Repeated measures ANOVA for knowledge us e 0.33 0.67 0.53 0.47 0 0.2 0.4 0.6 0.8 1 EG-HCG-H Knowledge use EG-H first; CG-H second (*) CG-H first; EG-H second (*) 0.53 0.47 0.29 0.71 0 0.2 0.4 0.6 0.8 1 EG-LCG-L Knowledge use EG-L first; CG-L second (*) CG-L first; EG-L second (*)
185 Overall, repeated measures ANOVA provided several i nteresting insights. Among those, one of the most salient was that when both governance mechanisms were set to the same (or a comparable) credibility condi tion, subjects were influenced by the recency effect, which inflated the effects of the s econd treatment in the sequence. If the credibility conditions of both mechanisms were set to high, subjects were more favorable toward the second page. On the other hand, if both governance mechanisms were set to low credibility condition, subjects perceptions of the credibility of the second mechanism were magnified again, resulting in an und ervaluation of the second treatment. Although these findings indicate the existence of o rder effects, they are still important, because the experimental design can be considered a good, if not perfect, representation of real world knowledge use situations. Since indi viduals retrieve knowledge from repositories in a sequential manner (i.e. one after another), the findings suggest that, when repositories have the same or a comparable lev el of credibility, individuals perceptions of the last piece of knowledge that the y are exposed to may override their perceptions of the previous knowledge assets they r etrieved. Assumptions In order to test the validity of the findings repor ted above, the assumptions of the techniques used in this study need to be validated. It has been acknowledged that PLS does not make any distributional assumptions unlike the covariance-based structural equation modeling (Barclay et al., 1995). However, the assumptions of repeated measures ANOVA have to be checked to ensure that th e findings are interpretable. The first assumption of repeated measures ANOVA is univ ariate and multivariate normality. Univariate normality was assessed by examining the skewness and kurtosis of each
186 measurement item. As previously discussed in Table 21, all measurement items were reasonably normal at the univariate level. The ske wness and kurtosis values of each item were within 2, which is a rule of thumb for normal ity (Hair et al., 2005). An exception was the expertise variable, which was highly skewed in favor of no expertise. However, this was expected, because the experiment was speci fically designed to minimize participants expertise in the experimental task. Further, expertise and involvement were excluded from the analysis as their moderating effe cts were controlled in the context of this study. Multivariate normality was assessed on the basis of univariate normality. It has been acknowledged in the literature that no techniq ue can sufficiently assess multivariate normality (Bentler and Chou, 1987). However, resea rchers argue that there are techniques that help infer multivariate normality or test it partially (Jres kog, 1993). One such technique relies on univariate assumption, and suggests that normality at the univariate level is a necessary condition for multi variate normality. Although univariate normality does not guarantee multivariate normality a non-normal univariate distribution is sufficient to infer lack of multivariate normali ty. Since the measurement items had acceptable univariate distributions, this study inf ers that the data also exhibit sufficient multivariate normality. It is also important to no te that even if there are deviations from multivariate normality, ANOVA is robust with respec t to normality. The second assumption of repeated measures ANOVA co ncerns the homogeneity of covariances. The findings of repeated measures ANOVA are based on the assumption that the covariance matrix of the dependent variabl es is the same for between-subject effects. The Boxs test of homogeneity enables to check this assumption, where a
187 significant test statistic indicates that the homog eneity of covariances is not equal. The pvalues of this test are presented in Table 29 for e ach group. It is worth mentioning that the Boxs test is applicable for only between-subje ct comparisons. The analysis conducted in this essay did not examine between-sub ject effects as those findings were non-interpretable. However, the use of counterbala nced design enabled to examine between-subject effects in a single group, and thus calculate the related test statistic. Therefore, the statistics reported in the table are computed separately for each group. Group 1 Group 2 Group 3 Group 4 Credibility of governance mechanism 0.06 0.685 0.073 0.543 Source credibility 0.001 0.526 0.411 0.472 Quality 0.710 0.126 0.709 0.374 Intention 0.942 0.071 0.072 0.483 Notes : 1) The Box test cannot be computed for the knowledg e use construct 2) Bold-faced values represent significant values a t =0.05 Table 29. P-values of Boxs homogeneity of covarian ces test As seen in the table, the p-value of source credibi lity in Group 1 was significant at an alpha level of 0.05, suggesting that the homogen eity of covariances was not equal for this construct. Therefore, the findings in this gr oup concerning source credibility need to be interpreted cautiously. The test also showed th at there were other p-values that were close to the cut-off value of 0.05. For example, t he test of the credibility of governance mechanism construct in Group 1 and Group 3, and the test of the intention construct in Group 2 and Group 3 were close to the cut-off alpha although they are were not considered significant. Therefore, caution needs t o be taken in interpreting the corresponding findings.
188 The third assumption of repeated measured ANOVA is sphericity, which suggests that in order for the findings to be interpretable the covariance matrix formed during the analysis should be in circular form. The test of s phericity is conducted using Mauchlys test. However, when the dependent variables have o nly two levels (which is the case in this study), Mauchlys test statistic cannot be com puted. This is because, the covariance matrix does not have enough values to make comparis ons for sphericity. Therefore, the assumption of sphericity is not applicable in this study. The fourth assumption of repeated measures ANOVA is homogeneity of variances. This assumption is assessed using Leven es test, where a non-significant test statistic indicates homogeneity of variances. The p-value of the test statistic for each variable in each group is presented in Table 30. A s seen in the table, all variances were homogeneous except the source credibility construct in the second group. Therefore, interpretations of the findings concerning source c redibility in this group require further caution. Group 1 Group 2 Group 3 Group 4 T1 T2 T1 T2 T1 T2 T1 T2 Credibility of governance mechanism 0.13 0.62 0.11 0.97 0.10 0.87 0.13 0.45 Source credibility 0.22 0.53 0.55 0.02 0.22 0.70 0.39 0.29 Quality 0.42 0.47 0.08 0.09 0.15 0.77 0.07 0.07 Intention 0.79 0.45 0.30 0.73 0.09 0.39 0.50 0.58 Knowledge use 0.14 0.14 0.29 0.29 0.70 0.70 0.07 0. 07 Notes : 1) Bold-faced values represent significant values a t =0.05 2) T1: The first page provided to a subject in that group; T2: The second page provided to the same subject in the group. Table 30. P-values of Levenes homogeneity of varia nces test
189 Discussion Key findings The goal of this essay was to understand the nature and the effect of factors that influenced individuals use of knowledge from exper tand community-governed repositories. The specific research questions of i nterest were: (a) what factors influence individuals use of knowledge from expertand comm unity-governed repositories; and (b) how? To answer these questions, this study ado pted a positivist perspective and employed the elaboration likelihood model (ELM) to design an experiment. As a theory of attitude formation, ELM suggested th at individuals relied on central and peripheral routes contingent upon their elaboration likelihood for using knowledge from repositories. Based on prior litera ture, the peripheral route was operationalized using source credibility, the centr al route using knowledge quality, and the elaboration likelihood using individuals exper tise and involvement in the experimental task. Additionally, a new peripheral route construct, namely the credibility of governance mechanism, was added into the researc h model to account for the variation in knowledge use due to the existence of governance mechanisms. The proposed model also theorized that the central route did not work in isolation, but was influenced by the peripheral route. Therefore, the source credibilit y and the credibility of governance mechanism constructs were hypothesized to bias indi viduals perceptions of knowledge quality. Therefore, a total of three hypotheses we re tested in this study, two for the effects of the credibility of governance mechanism on knowledge quality and intention, and one for the effect of intention on actual knowl edge use. The experiment to test these
190 hypotheses was designed such that only the credibil ity of governance mechanisms were manipulated, while keeping the other constructs con stant across all experimental groups. Testing these three hypotheses on the data collecte d from undergraduate students revealed that the hypothesized relationships were v alid for both expertand communitygovernance. In line with existing research, indivi duals intention to use knowledge was positively related to their knowledge use from both the expertand the communitygoverned page, as theorized in H1. The credibility of governance mechanism, the new peripheral route proposed in this study, positively influenced individuals intentions to use knowledge as well as their quality perceptions, supporting H2 and H3 respectively. Following the hypotheses testing, a post-hoc analys is was conducted using repeated measures ANOVA to compare individuals per ceptions across the two governance mechanisms examined in this study. The analysis focused on within-subject comparisons in all four groups. No between-subject comparisons were made, since the corresponding findings were not interpretable. The findings for those groups, in which subjects were exposed to one high credibility and o ne low credibility mechanism, were as expected, as individuals had more favorable percept ions toward the governance mechanism that was set to the high credibility cond ition (regardless of whether the mechanism was expertor community-governance). Ho wever, interesting findings were observed for the groups that received both governan ce mechanisms with high (or low) credibility conditions simultaneously. Concerning the credibility of governance mechanism, subjects perceived expert-governance to be more credible than communitygovernance when both mechanisms were set to high cr edibility condition; and perceived expert-governance to be less credible than communit y-governance when both
191 mechanisms were set to low credibility condition. The comparison concerning knowledge quality showed that subjects perceived th e quality of the community-governed page as being higher than that of the expert-govern ed page, when the two governance mechanisms were set to high credibility condition. This indicated the possibility of subjects showing greater appreciation for the coll ective effort afforded by the community in governing knowledge assets. On the ot her hand, when both governance mechanisms were set to low credibility condition, s ubjects were influenced by the recency effect, where they perceived the quality of the second page as being lower in quality. The recency effect also played a role in determining subjects intention to use knowledge and their actual use of knowledge. Accor dingly, the mean intention score and the knowledge use measure were higher for the secon d Web page used in the experiment when both governance mechanisms were set to high cr edibility condition. However, when both mechanisms were set to low credibility co ndition, the mean intention score and knowledge use measure were less favorable for t he second page. This indicated that when governance mechanisms had comparable levels of credibility, individuals were more influenced by the last knowledge asset they we re exposed to. In high credibility condition, they perceived the knowledge asset as be ing more credible, and in low credibility condition, they perceived it as being l ess credible than earlier knowledge assets they received. Limitations of the study The findings reported above needs to be interpreted within the limitations of this study. First, the study used students as a substit ute for knowledge workers in the experiment. Although the experimental task was spe cifically chosen to make it relevant
192 for the student population and their knowledge use behaviors, caution needs to be taken in generalizing the findings of this study to organ izational settings. Future studies can strive to replicate or extend the experiment used i n this study using organizational knowledge workers and possibly using knowledge asse ts taken from the repositories of these workers. Second, the experiment was conducted online at the convenience of study participants. Therefore, it was possible for parti cipants to search for additional information on the Web about the experimental task, or interact with each other in answering questions. Although, this can be a threa t for internal validity, conducting the experiment online helped recruit more participants for the experiment, reducing the possible effects of such uncontrolled behavior. Fu ture studies can conduct the same or a similar experiment in a controlled setting, where p articipants do not have access to the Web or cannot interact with each other. Third, the analysis of order effects showed that su bjects were influenced by the order in which treatments (i.e., Web pages) were pr ovided to them. The responses provided for a specific sequence of treatments were significantly different from the responses provided for the reverse sequence of the same set of treatments, indicating the problem of carry-over effects. Although separate a nalyses were conducted for both treatments sequences used in the experiment, the or der effects poses a threat for the validity of findings reported in this study. There fore, interpretations of the findings need to be made cautiously, especially in generalizing t hem to different populations or to different types of knowledge assets.
193 Fourth, the experiment did not involve a control tr eatment that could act as a base level for making more meaningful comparisons. The inclusion of the control treatment would also be an anchor for subjects while respondi ng to the questions related to the manipulated treatments in the experiment. The curr ent design induces subjects to use the first treatment as an anchor in providing responses to the second treatment. This, in turn, introduces the order effects, since changing the or der of treatments changes the anchor as well. In order to reduce this confound, future res earch can first expose the subjects to a control treatment that represents a base level, and then expose them to the manipulated treatment (whether the high credibility or the low credibility governance mechanism). This may not only eliminate the problem of order ef fects, but also enable to make more meaningful comparisons both withinand between-sub jects. Theoretical implications This study has several theoretical implications. F irst, the findings demonstrate that when governance mechanisms are used to increas e knowledge quality in repositories, the existing theoretical models proposed in the lit erature may not adequately represent what transpires as individuals use knowledge from r epositories. Prior models, which are mostly informed by ELM, operationalize the peripher al and the central routes of cognition using source credibility and knowledge qu ality respectively to explain knowledge use. Therefore, the predominant assumpti on in the literature is that if individuals perceive knowledge source as credible o r knowledge as being high quality, the likelihood of knowledge use increases. However such an explanation may fall short of studying knowledge use when repositories are gov erned by mechanisms that increase the quality of knowledge they retain. As demonstra ted in this essay, the use of
194 governance mechanisms, which is becoming more preva lent for knowledge repositories, invokes a new peripheral route construct for explai ning knowledge use. Therefore, this study contributes to our current theoretical unders tanding of knowledge use by introducing a new peripheral route construct, namel y the credibility of a governance mechanism. This is important, as researchers need to account for contextual differences when a theory is borrowed from one context to be us ed in another. Since the use of governance mechanisms is becoming more common for k nowledge repositories (regardless of whether these repositories are on th e Web or in organizations), this extension is necessary to improve our understanding of knowledge use, and increase the explanatory power of existing theories. The second theoretical contribution of this study i nvolves the effect of the peripheral route on the central route in explaining knowledge use. Earlier studies that employ ELM suggest that central and peripheral rout es are independent of each other, forming judgments separately (Petty and Cacioppo, 1 986a; Petty and Cacioppo, 1986b). However, general dual-process theories, which opera te at a higher level of abstraction than ELM, suggest that it is not possible for centr al and peripheral routes to work in isolation (Slater and Rouner, 1996; Smith and DeCos ter, 2000). The two processes constantly interact with each other and influence o ne another preventing a single route to operate independent of the other. However, this in teraction has not garnered enough attention among KM researchers in explaining knowle dge use. Previous applications of ELM and its variants such as heuristic systematic modeling (HSM, Chaiken, 1980) hypothesize independent effects of central and peri pheral routes on knowledge use. This study, on the other hand, takes the dependency into account by theorizing the effects of
195 the peripheral route constructs on the central rout e construct. The positive and significant paths from the peripheral route constructs to the c entral route construct validate this argument, and indicate that using knowledge from re positories is more complex than it has originally been hypothesized by KM researchers. Specifically, cues about the knowledge source or the credibility of governance m echanism are likely to bias individuals perceptions of knowledge quality. The refore, even though two contributions have comparable levels of quality, individuals will have more favorable attitudes toward the one governed by a credible mechanism, or provid ed by a credible source. This extends the current applications of ELM in the cont ext of KM, and adds to our knowledge base that perceptions of knowledge qualit y are biased by peripheral factors. The third theoretical contribution of this study co ncerns the findings of the repeated measures ANOVA. One of the findings sugge sted by repeated measures ANOVA is that individuals intentions to use knowle dge and their actual use of knowledge are influenced by recency effects. There fore, when individuals retrieve different pieces of information from the Web or fro m their organizations knowledge repositories, and if these pieces of information ha ve comparable levels of credibility, individuals are more likely to use the one that is retrieved last. To the best of our knowledge, current theoretical frameworks used in t he domain of KM do not take this temporality into consideration. This is especially important for developing a grand theory of knowledge use, in which the addition of s uch contingent factors can increase the explanatory power. This study has important research implications as w ell. To the best of our knowledge, this is the first study that examines ho w individuals use knowledge from
196 repositories that are governed by different governa nce mechanisms. Previous studies in the literature neither mention governance mechanism s nor investigate how they influence knowledge use behaviors. In this sense, this resea rch addresses a gap in the literature, and is expected to stimulate research on a couple o f fronts. First, this study argues that when repositories emp loy governance mechanisms to increase knowledge quality, credibility of the gove rnance mechanisms become a salient antecedent of knowledge use from these repositories In doing so, this study assumes that individuals credibility perceptions are their over all evaluation of the different aspects of governance mechanisms. For instance, in the case o f community-governance, credibility perceptions are based on the extent of the number o f edits, the number and the intensity of ratings, and the credibility of community member s. Although such an assumption is not unreasonable, further research can examine the effects of the different aspects of governance mechanisms individually without aggregat ing them under the umbrella of the credibility construct. For example, in community-g overnance, researchers can introduce new peripheral route constructs concerning number o f edits, number of ratings, quality of ratings, comments, revisions, credibility of commun ity, etc. to open up the credibility construct and understand the most salient aspects o f community-governance in explaining intentions and knowledge use. This can also increa se the explanatory power of the models proposed in this study and provide more insi ghts about how governance mechanisms influence knowledge use. Such an invest igation may not only further theory development efforts, but also provide guidance for designing new technologies and new governance mechanisms for knowledge management.
197 Second, the experimental design used in this study controls individuals elaboration likelihood and forces them to use the p eripheral routes in making judgments about the information provided to them. However, k nowledge users also use the central route besides the peripheral route as they make jud gments about the information they would like to use. Therefore, future research can investigate the proposed model in settings where knowledge users can use both periphe ral and central routes contingent upon their elaboration likelihood. This may provid e further insights about how and when governance mechanisms play a role in using knowledg e from repositories. However, such an investigation requires elaboration likeliho od to vary, allowing users to choose the route that best fits their decision making ability in a given context. Since elaboration likelihood is a context-dependent construct, resear chers may need to develop more complex experiments in different contexts. Develop ing such experiments inflate the number of manipulations and experimental conditions that need to be created, and thus increase the sample size requirements. In order to eliminate such logistical problems, future research can use agent-based modeling to sim ulate those conditions, and investigate the salience of governance mechanisms. Third, the new peripheral route construct developed in this study is hypothesized to have two dimensions: credibility of the governor s, and the credibility of the page as a result of the governance processes. While the form er concerns the trustworthiness, reliability, expertise, and knowledge of experts (i n expert-governance) or community (in community-governance), the latter involves the trus tworthiness and reliability of the knowledge asset resulting from the governance proce sses. These two dimensions can vary independent of each other as knowledge users c an perceive experts or community
198 members as credible but the knowledge asset as less credible (and vice versa). This study manipulated these two concepts simultaneously in or der to eliminate any measurement related confounds. Future research can manipulate these two sub-dimensions independently, and try to understand the dimension that is most salient in influencing individuals intentions to use knowledge. Practical implications This study has several practical implications as we ll. First and foremost, this study demonstrates that governance mechanisms that are employed for knowledge repositories influences individuals knowledge use behaviors. Organizations make significant investments in knowledge repositories t o create organizational memory, document salient processes and procedures, and help individuals inside or outside organizational boundaries reuse the knowledge store d in these repositories. However, if these repositories do not store high quality knowle dge, their likelihood of being used by organizational stakeholders decreases. Therefore, in addition to investing in technology, more organizations are starting to invest in govern ance mechanisms (such as expertand community-governance) to improve the quality of kno wledge stored in repositories. The credibility of such mechanisms, as demonstrated in this study, influences individuals intentions to use knowledge, which ultimately affec ts actual knowledge use. Practitioners can leverage this finding to increase their stakeholders use of knowledge from their repositories in two ways: (1) by ensuring that the governance functions used to increase knowledge quality are ro bust, effective, and executed appropriately so that they are able to increase qua lity of knowledge stored in repositories; (2) by making sure that the individuals (i.e., expe rts or community members) who
199 execute the governance processes are credible. By addressing these two issues, practitioners can increase the credibility of the g overnance mechanisms used for their knowledge repositories, which in turn influences qu ality perceptions as well as intentions. Therefore, organizations that have public repositor ies on the Web can attract more users (and thus more traffic) to their sites, and those t hat use repositories for organizational knowledge management can increase the extent of kno wledge transfer among organizational members (and thus enjoy higher level s of efficiency and effectiveness). A second implication of this study is that credibil ity of governance mechanism influences individuals perceptions of knowledge qu ality. This indicates that if individuals encounter knowledge assets that serve t he same need, they can perceive the one that employs a credible governance mechanism as being higher in quality. This can be true even if the content quality of the two know ledge assets do not differ significantly. Since individuals are more likely to use knowledge if they have favorable perceptions about its quality (Zack, 1999), organizations can f urther boosts knowledge use from repositories by implementing a credible governance mechanisms. Third, findings concerning the effects of the credi bility of governance mechanism have implications for the design of knowledge repos itories. Both governance mechanisms (but especially expert-governance) are s usceptible to agency problems, where knowledge users may not be aware of the types or the quality of governance functions executed on knowledge assets. If this in formation is not conveyed to knowledge users appropriately, users may perceive t he credibility of a related governance mechanism less favorable than it is, which may infl uence the use of knowledge assets stored in the repository. For this reason, practit ioners may need to make sure that
200 repositories are designed to present meta-data to k nowledge users about the types and the quality of governance functions executed by experts or community members. The fourth implication of the study concerns indivi duals perceptions of the credibility of expertand community-governance. F indings suggest that individuals may perceive expert-governance as being more credible t han community-governance even though both mechanisms are equally credible. This may indicate that individuals may be predisposed to expert-governance since it is the mo st commonly used mechanism for increasing knowledge quality for centuries (Kronick 1990). Therefore, expertgovernance can be perceived as being more credible than community-governance regardless. However, this differential may erode d ue to the latest developments in technology that aim to harness the collective power of individuals in solving challenging problems. Especially, the trend in experimenting w ith technologies such as wikis and discussion forums can demonstrate the power of comm unity-governance compared to expert-governance, and can dethrone the dominance o f expert-governance in the future.
201 CONCLUSION The goals of this dissertation were to set the conc eptual foundations of the governance concept for increasing knowledge quality in electronic repositories, understand the aspects of two commonly used governa nce mechanisms that contribute to knowledge quality, and examine how individuals made contributions to and used knowledge from repositories in the existence of the se two mechanisms. The dissertation tried to achieve these goals in three related essay s. The first essay developed the concept of governance by drawing upon the governance litera ture in sociology. After identifying four different governance mechanisms, it focused on expertand community-governance in detail, and examined whether these two mechanism s increased quality of knowledge in repositories, and why or why not. Using an interpr etive paradigm, this essay conducted qualitative research by collecting empirical data f rom professionals who used expertand community-governance in their firms. The findings not only identified the aspects of both governance mechanisms that contributed to know ledge quality, but also provided additional insights about how individuals perceived these two governance mechanisms in organizational settings. This essay informs the se cond and third essays of the concept of governance, and paves the way for investigating the knowledge contribution and knowledge use behaviors in the existence of expertand community-governance. The findings of this essay also inform the third essay, as some of the hypotheses tested in the third essay draw upon the findings reported in this essay.
202 The second essay concerned the factors that were sa lient for contributing to repositories governed with the two mechanisms conce ptualized in the first essay. The specific research question examined in this essay w as: what factors influence individuals to make voluntary contributions to expertand comm unity-governed repositories? This essay examined this research question in two differ ent contexts, one in which there was only one type of repository in use (either expert-g overned or community-governed), and another in which the two types of repositories were used simultaneously. Similar to the first essay, this essay adopted an interpretive par adigm and conducted qualitative research by collecting empirical data from professionals who used expertand communitygoverned repositories in both contexts. The findin gs revealed important insights for theory and practice. Especially, the factors that were salient for explaining contribution behaviors when the two repositories existed simulta neously not only laid the groundwork for a theory of choice, but also provided insights about the different uses of expertand community-governed repositories. The third essay concerned the use of knowledge from repositories when they employed expertand community-governance as a mean s to increase knowledge quality. The research question of interest to this essay was : (a) what factors influence individuals use of knowledge from expertand community-governe d repositories; and (b) how? Unlike the previous two essays, this essay adopted a positivist paradigm, and drew upon the elaboration likelihood model to propose a resea rch model about the salience of the credibility of governance mechanisms during knowled ge use. Specifically, it hypothesized that when governance mechanisms were u sed to increase knowledge quality in repositories, the credibility of those governanc e mechanisms influenced individuals
203 perceptions of quality and intentions to use knowle dge, which ultimately determined their knowledge use. Using a repeated measures experimen t, this essay provided support for the hypothesized relationships, and suggested that credibility of governance mechanisms was salient in explaining knowledge use. This essa y also conducted a post-hoc analysis using repeated measures ANOVA to compare individual s perceptions of the two governance mechanisms for different credibility lev els. An interesting and unexpected finding was that individuals had more favorable per ceptions for the last knowledge asset they were exposed to, if the credibility of the gov ernance mechanisms used for those knowledge assets were comparable. The three essays of this dissertation contribute to our current theoretical knowledge in different ways. The first essay sugge sts propositions about the different aspects of expertand community-governance that co ntribute to knowledge quality, the second essay develops two theoretical models to exp lain contribution behaviors for two different contexts, and the third essay extends the elaboration likelihood model to explain knowledge use from expertand community-governed r epositories. Overall, the findings reported in this dissertation bring KM researchers one step closer to developing theories for governance mechanisms, knowledge contribution b ehaviors, and knowledge use. All three essays emphasize the need to incorporate the effects of governance mechanisms into our existing knowledge to develop new or exten d existing theories. The three essays of the dissertation also make impo rtant practical contributions. The first essay provides guidance to practitioners on how to instantiate effective governance mechanisms to increase the quality of kn owledge in repositories, and how to reduce the agency problem between governors and kno wledge users through technology
204 design. The second essay provides suggestions abou t how to motivate organizational members to make more contributions to expertand c ommunity-governed repositories, and sheds light on why governance mechanisms matter if individuals are given a choice. The third essay highlights the importance of the cr edibility of governance mechanisms during knowledge use, and shows how credibility inf luences individuals perceptions of knowledge quality and their intention to use knowle dge. This dissertation has important research implicatio ns as well. The concept of governance the underlying theme of this dissertat ion provides many opportunities to refine our existing understanding of KM theories an d develop new ones. It also informs design science researchers of a new distinction bet ween KM technologies, and paves the way for the development and evaluation of various t echnological designs. Considering the different types of opportunities provided by th e governance concept, more research is needed to understand how governance mechanisms impa ct what we already know, and how they can inform the field of IS.
205 REFERENCES Adler, P.S., and Borys, B. 1996. "Two Types of Bure aucracy: Enabling and Coercive," Administrative Science Quarterly (41:1), pp 61-89. Ajzen, I. 2002. "Constructing a Tpb Questionnaire." Retrieved June 19th, 2009, from http://people.umass.edu/aizen/pdf/tpb.measurement.p df Ajzen, I., and Fishbein, M. 1980. Understanding Attitudes and Predicting Social Behavior Englewood Cliffs, NJ: Prentice-Hall. Alavi, M., Kayworth, T.R., and Leidner, D.E. 2006. "An Empirical Examination of the Influence of Organizational Culture on Knowledge Ma nagement Practices," Journal of Management Information Systems (22:3), pp 191-224. Alavi, M., and Leidner, D.E. 2001. "Review: Knowled ge Management and Knowledge Management Systems: Conceptual Foundations and Rese arch Issues," MIS Quarterly (25:1), pp 107-136. Amabile, T.M. 1993. "Motivational Synergy: Toward N ew Conceptualizations of Intrinsic and Extrinsic Motivation in the Workplace ," Human Resource Management Review (3:3), pp 185-201. Asch, S.E. 1946. "Forming Impressions of Personalit y," Journal of Abnormal and Social Psychology (41:3), pp 258-290. Axelrod, R.M. 1984. The Evolution of Cooperation New York: Basic Books. Ba, S., Statlaert, J., and Whinston, A.B. 2001. "Re search Commentary: Introducing a Third Dimension in Information Systems Design-the C ase for Incentive Alignment," Information Systems Research (12:3), pp 225-239. Bagozzi, R.P., and Dholakia, U.M. 2002. "Intentiona l Social Action in Virtual Communities," Journal of Interactive Marketing (16:2), pp 2-21. Bagozzi, R.P., and Phillips, L.W. 1982. "Representi ng and Testing Organizational Theories: A Holistic Construal," Administrative Science Quarterly (27:3), pp 459489. Bandura, A. 1986. Social Foundations of Thought and Action : A Social Cognitive Theory Englewood Cliffs, N.J.: Prentice-Hall.
206 Barclay, D., Higgins, C., and Thompson, R. 1995. "T he Partial Least Squares (Pls) Approach to Causal Modeling: Personal Computer Adop tion and Use as an Illustration," Technology Studies (2:2), pp 285-324. Bentler, P.M., and Chou, C. 1987. "Practical Issues in Structural Modeling," Sociological Methods & Research (16:1), pp 78117. Bhattacherjee, A., and Sanford, C. 2006. "Influence Process for Information Technology Acceptance: An Elaboration Likelihood Model," MIS Quarterly (30:4), pp 805825. Bock, G.-W., Zmud, R.W., Kim, Y.-G., and Lee, J.-N. 2005. "Behavioral Intention Formation in Knowledge Sharing: Examining the Roles of Extrinsic Motivators, Social-Psychological Forces, and Organizational Cli mate," MIS Quarterly (29:1), pp 87-111. Bowles, S., and Gintis, H. 2002. "Social Capital an d Community Governance," The Economic Journal (112), pp 419-436. Bryson, J.M., and Crosby, B.C. 1993. "Policy Planni ng and the Design of Forums, Arenas and Courts," in: Public Management, B. Bozeman (ed.). San Francisco, CA: Jossey Bass. Caplan, N.S. 1975. The Use of Social Science Knowledge in Policy Decis ions at the National Level: A Report to Respondents Ann Arbor, MI: Institute for Social Research, University of Michigan. Centefelli, R.T. 2004. "Inhibitors and Enablers as Dual Factor Concepts in Technology Usage," Journal of the Association for Information Systems (5:11-12), pp 472492. Chaiken, S. 1980. "Heuristic Versus Systematic Info rmation Processing and the Use of Source Versus Message Cues in Persuasion," Journal of Personality and Social Psychology (39:5), pp 752-766. Chaiken, S., Liberman, A., and Eagly, A.H. 1989. "H euristic and Systematic Information Processing within and Beyond the Persuasion Context ," in: Unintended Thought, J.S. Uleman and J.A. Bargh (eds.). New York: Guilfo rd Press, pp. 212252. Charmaz, K. 2000. "Grounded Theory: Objectivist and Constructivist Methods," in: Handbook of Qualitative Research, N.K. Denzin and Y.S. Lincoln (eds.). Thousand Oaks, CA: Sage, pp. 509535. Chen, I.Y.L. 2007. "The Factors Influencing Members Continuance Intentions in Professional Virtual Communities a Longitudinal Stu dy," Journal of Information Science (33:4), pp 451-467.
207 Chiu, C.M., Hsu, M.H., and Wang, E.T.G. 2006. "Unde rstanding Knowledge Sharing in Virtual Communities: An Integration of Social Capit al and Social Cognitive Theories," Decision Support Systems (42:3), pp 1872-1888. Choi, C.J., Cheng, P., Hilton, B., and Russell, E. 2005. "Knowledge Governance," Journal of Knowledge Management (9:6), pp 67-75. Choudhury, V., and Sabherwal, R. 2001. "Information Technology and the Boundary of the Firm: A Knowledge-Based Perspective," Redefining the Organizational Roles of Information Technology in the Information Age, R.W. Zmud, A. Segars and J. Sampler (eds.), Cincinnati, OH: Pinnaflex. Chourides, P., Longbottom, D., and Murphy, W. 2003. "Excellence in Knowledge Management: An Empirical Study to Identify Critical Factors and Performance Measures," Measuring Business Excellence (7:2), pp 29-45. Chow, W.S., and Chan, L.S. 2008. "Social Network, S ocial Trust and Shared Goals in Organizational Knowledge Sharing," Information & Management (45), pp 458465. Cohen, J., and Cohen, P. 1983. Applied Multiple Regression/Correlation Analysis fo r the Behavioral Sciences (Second ed.). Hillsdale, New Jersey: Lawrence Erl baum Associates. Constant, D., Kiesler, S., and Sproull, L. 1994. "W hat's Mine Is Ours, or Is It? A Study of Attitudes About Information Sharing," Information Systems Research (5:4), pp 400-421. Cosley, D., Frankowski, D., Kiesler, S., Terveen, L ., and Riedl, J. 2005. "How Oversight Improves Member-Maintained Communities," in: The SIGCHI conference on Human factors in computing systems Portland, Oregon, USA: ACM. Cummings, J.N., Sproull, L., and Kiesler, S.B. 2002 "Beyond Hearing: Where RealWorld and Online Support Meet," Group Dynamics: Theroy, Research, and Practice (6:1), pp 78-88. Davenport, T.H. 1997. Information Ecology Oxford, England: Oxford University Press. Davenport, T.H., Eccles, R.G., and Prusak, L. 1992. "Information Politics," Sloan Management Review (34:1), pp 53-65. Davenport, T.H., Prusak, L., and Strong, B. 2008. Business Insight (a Special Report): Organization; Putting Ideas to Work: Knowledge Mana gement Can Make a Difference but It Needs to Be More Pragmatic," in : Wall Street Journal (Eastern Edition) p. R.11. Deci, E.L., and Ryan, R.M. 1985. Intrinsic Motivation and Self-Determination in Huma n Behavior New York: Plenum.
208 DeLone, W.H., and McLean, E.R. 1992. "Information S ystems Success: The Quest for the Dependent Variable," Information Systems Research (3:1), pp 60-95. Dholakia, U.M., Bagozzi, R.P., and Pearo, L.K. 2004 "A Social Influence Model of Consumer Participation in Network-and Small-Group-B ased Virtual Communities," International Journal of Research in Marketing (21:3), pp 241263. Dijkstra, J.J. 1995. "The Influence of an Expert Sy stem on the Users View: How to Fool a Lawyer," New Review of Applied Expert Systems (1), pp 123-138. Dijkstra, J.J. 1999. "User Agreement with Incorrect Expert System Advice," Behavior & Information Technology (18:6), pp 399-411. Dijkstra, J.J., Liebrand, W.B.G., and Timminga, E. 1998. "Persuasiveness of Expert Systems," Behaviour & Information Technology (17:3), pp 155-163. Dixon, N.M. 2000. Common Knowledge: How Companies Thrive by Sharing W hat They Know Boston: Harvard Business School Press. Dretske, F. 1981. Knowledge and the Flow of Information Cambridge, MA: MIT Press. Drucker, P. 1988. "The Coming of the New Organizati on," Harvard Business Review (66), pp 45-53. Fadel, K.J., Durcikova, A., and Cha, H.S. 2008. "El aboration Likelihood in Knowledge Management: A Model and Experimental Test," The 41st Hawaii International Conference on System Sciences Hawaii, pp. 359-368. Festinger, L. 1957. A Theory of Cognitive Dissonance Stanford, CA: Stanford University Press. Flanagan, J.C. 1954. "The Critical Incident Techniq ue," Psychological bulletin (51:4), pp 327-358. Fornell, C., and Larcker, D.F. 1981. "Evaluating St ructural Equations with Unobservable Variables and Measurement Error," Journal of Marketing Research (18:1), pp 3950. Foss, N.J. 2007. "The Emerging Knowledge Governance Approach: Challenges and Characteristics," Organization (14:1), pp 2952. Gawronski, B., and Bodenhausen, G.V. 2006. "Associa tive and Propositional Processes in Evaluation: An Integrative Review of Implicit an d Explicit Attitude Change," Psychological Bulletin (132:5), pp 692-731.
209 Glaser, B.G. 1978. Theoretical Sensitivity: Advances in the Methodolog y of Grounded Theory San Francisco, CA: Sociology Press. Glaser, B.G., and Strauss, A.L. 1967. The Discovery of Grounded Theory; Strategies for Qualitative Research Chicago,: Aldine. Goodhue, D.L. 1995. "Understanding User Evaluations of Information Systems," Management Science (41:12), pp 1827-1844. Gottschalk, P. 2000. "Knowledge Management Systems: A Comparison of Law Firms and Consulting Firms," Informing Science (3:3), pp 117-124. Grandori, A. 2001. "Neither Hierarchy nor Identity: Knowledge-Governance Mechanisms and the Theory of the Firm," Journal of Management and Governance (5), pp 381-399. Grant, D.A. 1948. "The Latin Square Principle in th e Design and Analysis of Psychological Experiments," Psychological bulletin (45:5), pp 427-442. Hackman, J.B., and Morris, C.G. 1975. "Group Tasks, Group Interaction Process, and Group Performance Effectiveness: A Review and Propo sed Integration," in: Advances in Experimental Social Psychology, L. Berkowitz (ed.). San Diego, CA.: Academic Press. Hair, J.F., Black, B., Babin, B., Anderson, R.E., a nd Tatham, R.L. 2005. Multivariate Data Analysis (6 ed.). New Jersey: Prentice Hall. Hansen, M.T., Nohria, N., and Tierney, T. 1999. "Wh ats Your Strategy for Managing Knowledge?," Harvard Business Review (77:2), pp 106-116. He, W., and Wei, K.-K. 2009. "What Drives Continued Knowledge Sharing? An Investigation of Knowledge-Contribution and -Seekin g Beliefs," Decision Support Systems (46:4), pp 826-838. Herzberg, F., Mausner, B., and Snyderman, B. 1959. The Motivation to Work New York: John Wiley and Sons. Holzner, B., and Marx, J.H. 1979. Knowledge Application: The Knowledge System in Society Boston: Allyn-Bacon. Huber, G.P. 1991. "Organizational Learning: The Con tributing Processes and the Literatures," Organization Science (2:1), pp 88-115. ITGI. 2003. "Board Briefing on It Governance." Re trieved April, 06, 2010, from http://www.itgi.org/AMTemplate.cfm?Section=Board_Br iefing_on_IT_Governan ce&Template=/ContentManagement/ContentDisplay.cfm&C ontentID=39649 Janis, I.L. 1982. Groupthink Boston, MA: Houghton-Mifflin.
210 Jarke, M., and Vassiliou, Y. 1997. "Data Warehouse Quality: A Review of the Dwq Project," Proceedings of the Conference on Information Qualit y Cambridge, MA, pp. 299313. Jarvenpaaa, S.L., and Staples, D.S. 2000. "The Use of Collaborative Electronic Media for Information Sharing: An Exploratory Study of Determ inants," Journal of Strategic Information Systems (9), pp 129-154. Jreskog, K.G. 1993. "Testing Structural Equation M odels," in: Testing Structural Equation Models, K.A. Bollen and J.S. Long (eds.). Newbury Park, CA : Sage. Kalman, M.E., Monge, P., Fulk, J., and Heino, R. 20 02. "Motivations to Resolve Communication Dilemmas in Database-Mediated Collabo ration," Communication Research (29:2), pp 125-154. Kankanhalli, A., Tan, B.C.Y., and Wei, K.-K. 2005. "Contributing to Electronic Knowledge Repositories: An Empirical Investigation, MIS Quarterly (29 1), pp 113-143. Kaplan, B., and Maxwell, J.A. 1994. "Qualitative Re search Methods for Evaluating Computer Information Systems," in: Evaluating Health Care Information Systems: Methods and Applications, J.G. Anderson, C.E. Aydin and S.J. Jay (eds.). Thousand Oaks, CA: Sage, pp. 45-68. Keppel, G. 1991. Design and Analysis: A Researchers Handbook Englewood Cliffs, NJ: Prentice Hall. Koh, J., Kim, Y.G., Butler, B., and Bock, G.W. 2007 "Encouraging Participation in Virtual Communities," Communications of the ACM (50:2), pp 69-73. Kooiman, J. 1999. "Social-Political Governance," Public Management Review (1:1), pp 67-92. Kooiman, J., and Bavinck, M. 2005. "The Governance Perspective," in: Fish for Life: Interactive Governance for Fisheries, J. Kooiman, S. Jentoft, R. Pullin and M. Bavinck (eds.). Amsterdam University Press, pp. 1124. Krafft, J., and Ravix, J.-L. 2008. "Corporate Gover nance and the Governance of Knowledge: Rethinking the Relationship in Terms of Corporate Coherehce," Economics of Innovation and New Technology (17:1-2), pp 79-96. Kronick, D.A. 1990. "Peer Review in 18th-Century Sc ientific Journalism," The Journal of American Medical Association (263:10), pp 1321-1322. Lank, E., Randell-Khan, J., Rosenbaum, S., and Tate O. 2008. "Herding Cats: Choosing a Governance Structure for Your Communities of Prac tice," Journal of Change Management (8:2), pp 101-109.
211 Larsen, J. 1980. "Knowledge Utilization: What Is It ?," Knowledge, Creation, Diffusion, Utilization (3), pp 421-442. Lee, Y.W., Strong, D.M., Kahn, B.K., and Wang, R.Y. 2002. "Aimq: A Methodology for Information Quality Assessment," Information & Management (40:2), pp 133146. Lin, T.-C., and Huang, C.-C. 2008. "Understanding K nowledge Management System Usage Antecedents: An Integration of Social Cogniti ve Theory and Task Technology Fit," Information & Management (45 6), pp 410-417. Lincoln, Y., and Guba, E. 1985. Naturalistic Inquiry Beverly Hills, CA: Sage. Machlup, F. 1980. Knowledge: Its Creation, Distribution, and Econo mic Significance Princeton, NJ: Princeton University Press. Mak, B., Schmitt, B.H., and Lyytinen, K. 1997. "Use r Participation in Knowledge Update of Expert Systems," Information & Management Science (32:2), pp 55-63. Markus, M.L. 2001. "Toward a Theory of Knowledge Re use: Types of Knowledge Reuse Situations and Factors in Reuse Success," Journal of Management Information Systems (18:1), pp 57-93. Martin, P.Y., and Turner, B.A. 1986. "Grounded Theo ry and Organizational Research," The Journal of Applied Behavioral Science (22:2), pp 141-157. McAdam, R., and McCreedy, S. 1999. "A Critical Revi ew of Knowledge Management Models," The Learning Organization (6:3), pp 91-100. McKinsey. 2008. "Building the Web 2.0 Enterprise: M ckinsey Global Survey Results," in: The McKinsey Quarterly McNeely, B.L., and Meglino, B.M. 1994. "The Role of Dispositional and Situational Antecedents in Prosocial Organizational Behavior: A n Examination of the Intended Beneficiaries of Prosocial Behavior," Journal of Applied Psychology (79:6), pp 836-844. Moorman, C., and Miner, A.S. 1998. "Organizational Improvisation and Organizational Memory," Academy of Management Review (23:4), pp 698-723. Mu, J., Peng, G., and Love, E. 2008. "Interfirm Net works, Social Capital, and Knowledge Flow," Journal of Knowledge Management (12:4), pp 86-100. Myers, M.D. 1997. "Qualitative Research in Informat ion Systems," MIS Quarterly (21:2), pp 241-242.
212 Neus, A. 2001. "Managing Information Quality in Vir tual Communities of Practice," 6th International Conference on Information Quality, E. Pierce and R. Katz-Haas (eds.), MIT, Boston, MA: Sloan School of Management Neus, A., and Scherf, P. 2004. "Managing Knowledge Peer-to-Peer Concepts and Challenges," International Journal of Internet and Enterprise Ma nagement (2:1), pp 96-111. Nevo, D., Benbasat, I., and Wand, Y. 2009. "Knowled ge Management Who Knows What?," in: Wall Street Journal (Eastern Edition) p. R.4. Nonaka, I. 1994. "A Dynamic Theory of Organizationa l Knowledge Creation," Organization Science (5:1), pp 14-37. O'Brien, R.G. 1979. "A General Anova Method for Rob ust Tests of Additive Models for Variances," Journal of the American Statistical Association (74:368), pp 877-880. Oh, C.H. 1997. "Issues for the New Thinking of Know ledge Utilization: Introductory Remarks," Knowledge, Technology, and Policy (10:3), pp 3-10. Penny, K.I. 1996. "Appropriate Critical Values When Testing for a Single Multivariate Outlier by Using the Mahalanobis Distance," Applied Statistics (45:1), pp 73-81. Pentland, B.T. 1995. "Information Systems and Organ izational Learning: The Social Epistemology of Organizational Knowledge Systems," Accounting, Management and Information Technologies (5:1), pp 1-21. Petty, R.E., and Cacioppo, J.T. 1986a. Communication and Persuasion: Central and Peripheral Routes to Attitude Change New York: Springer-Verlag. Petty, R.E., and Cacioppo, J.T. 1986b. "The Elabora tion Likelihood Model of Persuasion," Advances in experimental social psychology (19:1), pp 123-205. Petty, R.E., Cacioppo, J.T., and Schumann, D. 1983. "Central and Peripheral Routes to Advertising Effectiveness: The Moderating Role of I nvolvement," The Journal of Consumer Research (10:2), pp 135-146. Polanyi, M. 1958. Personal Knowledge Chicago: The University of Chicago Press. Pollatsek, A., and Well, A.D. 1995. "On the Use of Counterbalanced Designs in Cognitive Research: A Suggestion for a Better and M ore Powerful Analysis," Journal of Experimental Psychology: Learning, Memor y, and Cognition (21:3), pp 785-794. Pornpitakpan, C. 2004. "The Persuasiveness of Sourc e Credibility: A Critical Review of Five Decades' Evidence," Journal of Applied Social Psychology (34:2), pp 243281.
213 Quintas, P., Lefrere, P., and Jones, G. 1997. "Know ledge Management: A Strategic Agenda," Journal of Long Range Planning (30:3), pp 385-391. Reese, H.W. 1997. "Counterbalancing and Other Uses of Repeated-Measures LatinSquare Designs: Analyses and Interpretations," Journal of Experimental Child Psyhology (64), pp 137-158. Rich, R.F. 1975. "Selective Utilization of Social S cience Related Information by Federal Policy-Makers," Inquiry (12:3), pp 239 -245. Ringle, C.M., Wende, S., and Will, S. 2005. "Smartp ls 2.0 (M3) Beta," Hamburg, Germany, http://www.smartpls.de Schroeder, A., and Pauleen, D. 2007. "Km Governance : Investigating the Case of a Knowledge Intensive Research Organisation," Journal of Enterprise Information Management (20:4), pp 414-431. Schuler, D. 1994. "Community Networks: Building a N ew Participatory Medium," Communications of the ACM (37 1), pp 38-51. Slater, M.D., and Rouner, D. 1996. "How Message Eva luation and Source Attributes May Influence Credibility Assessment and Belief Cha nge," Journalism and Mass Communication Quarterly (73:4), pp 974-991. Sloman, S.A. 1996. "The Empirical Case for Two Syst ems of Reasoning," Psychological Bulletin (119:1), pp 3-22. Smith, E.R., and DeCoster, J. 2000. "Dual-Process M odels in Social and Cognitive Psychology: Conceptual Integration and Links to Und erlying Memory Systems," Personality and Social Psychology Review (4), pp 108-131. Strauss, A.L., and Corbin, J.M. 1998. Basics of Qualitative Research Techniques and Procedures for Developing Grounded Theory (2nd ed.). Thousand Oaks: Sage Publications. Streeck, W., and Schmitter, P.C. 1985. "Community, Market, State-and Associations? The Prospective Contribution of Interest Governance to Social Order," European Sociological Review (1:2), pp 119-138. Strong, B., Davenport, T.H., and Prusak, L. 2008. Organizational Governance of Knowledge and Learning," Knowledge and Process Management (15:2), pp 150157. Surowiecki, J. 2004. The Wisdom of Crowds New York, NY: Doubleday. Sussman, S.W., and Siegal, W.S. 2003. "Informationa l Influence in Organizations: An Integrated Approach to Knowledge Adoption," Information Systems Research (14:1), pp 47-65.
214 Swan, J., and Newell, S. 2000. "Linking Knowledge M anagement and Innovation," in: 8th European Conference on Information Systems, H.R. Hansen, M. Bichler and H. Mahrer (eds.). Vienna, Austria: Wirtschaftsunive rsitt Wien, pp. 591-598. Swan, J., Newell, S., Scarbrough, H., and Hislop, D 1999. "Knowledge Management and Innovation: Networks and Networking," Journal of Knowledge Management (3:4), pp 262-275. Tuomi, I. 1999. "Data Is More Than Knowledge: Impli cations of the Reversed Knowledge Hierarchy for Knowledge Management and Or ganizational Memory," in: The 32nd Hawaii International Conference on System Sciences Hawaii. Vance, D.M. 1997. "Information, Knowledge and Wisdo m: The Epistemic Hierarchy and Computer-Based Information System," Proceedings of the Third Americas Conference on Information Systems, B. Perkins and I. Vessey (eds.), Indianapolis. Venkatesh, V., Morris, M.G., Davis, G.B., and Davis F.D. 2003. "User Acceptance of Information Technology: Toward a Unified View," MIS Quarterly (27:3), pp 425478. Wang, R.Y., and Strong, D.M. 1996. "Beyond Accuracy : What Data Quality Means to Data Consumers," Journal of Management Information Systems (12:4), pp 5-34. Wasko, M.M., and Faraj, S. 2000. ""It Is What One D oes": Why People Participate and Help Others in Electronic Communities of Practice," Journal of Strategic Information Systems (9), pp 155-173. Wasko, M.M., and Faraj, S. 2005. "Why Should I Shar e? Examining Social Capital and Knowledge Contribution in Electronic Networks of Pr actice," MIS Quarterly (29:1), pp 3557. Weiss, C.H. 1979. "The Many Meanings of Research Ut ilization," Public Administration Review (39:5), pp 426-431. Wiig, K. 1997. "Knowledge Management: Where Did It Come from and Where Will It Go?," Expert Systems With Applications (13:1), pp 1-14. Wiig, K.M. 1995. Knowledge Management Methods: Practical Approaches to Managing Knowledge Arlington, TX: Schema Press. Winer, B.J., Brown, D.R., and Michels, K.M. 1991. Statistical Principles in Experimental Design New York: McGraw-Hill. Wynne, W.C., Barbara, L.M., and Peter, R.N. 2003. A Partial Least Squares Latent Variable Modeling Approach for Measuring Interactio n Effects: Results from a Monte Carlo Simulation Study and an ElectronicMail Emotion/Adoption Study," Information Systems Research (14:2), pp 189-217.
215 Zack, M.H. 1999. "Managing Codified Knowledge," Sloan Management Review (40:4), pp 45-58. Zaichkowsky, J.L. 1985. "Measuring the Involvement Construct," The Journal of Consumer Research (12:3), pp 341-352. Zander, U., and Kogut, B. 1995. "Knowledge and the Speed of the Transfer and Imitation of Organizational Capabilities: An Empirical Test," Organization Science (6:1), pp 76-92. Zhang, W., and Watts, S.A. 2008. "Capitalizing on C ontent: Information Adoption in Two Online Communities," Journal of the Association for Information Systems (9:2), pp 73-94. Zmud, R. 1978. "Concepts, Theories and Techniques: An Empirical Investigation of the Dimensionality of the Concept of Information," Decision Sciences (9:2), pp 187195. Zyngier, S., Burstein, F., and McKay, J. 2006. "The Role of Knowledge Management Governance in the Implementation of Strategy," in: The 39th Hawaii International Conference on System Sciences Hawaii.
Appendix A Figure A 217 Figure A 1. High credibility expert-governed page
Figure A 218 Figure A 2. Low credibility expert-governed page
Figure A 219 Figure A 3. High credibility community-governed page
Figure A 220 Figure A 4. Low credibility community-governed page
221 Appendix B Figure B 1. Instructions given to subjects
222 Appendix C Figure C 1. The link of the first treatment provi ded to subjects
Appendix D Figure D 1. Sample c 223 Sample c omprehension questions related to the governance mechanism related to the governance mechanism
Figure D 2. Sample c 224 Sample c omprehension questions related to the information o n omprehension questions related to the information o n a Web page
225 Appendix E Figure E 1. Measurement of knowledge use from the two pages
226 Appendix F Expert-governed Web page Notes: 1) Values without parentheses: subjects were given the expert-governed page first (EG1); Values with parentheses: subjects were given th e expert-governed page second (EG2). Figure F 1. Interaction effects model for the exp ert-governed page
227 Community-governed Web page Notes: 1) Values without parentheses: subjects were given the community-governed page second (CG2); Values with parentheses: subjects were give the community-governed page first (CG1). Figure F 2. Interaction effects model for the com munity-governed page
ABOUT THE AUTHOR Varol Kayhan received his Bachelors Degree in Mech anical Engineering from Middle East Technical University in Ankara, Turkey, in 1999. He worked at a multinational bank as a business analyst in Istanbu l, Turkey, until he entered the M.S. MIS program at the University of South Florida in 2 005. While pursuing his Masters degree, Varol applied to the Ph.D. program at the U niversity of South Florida and started to work toward his doctorate degree in addition to the Masters degree. He earned his Masters degree in 2008 during the second year of h is Ph.D. endeavor. While in the Ph.D. program, Varol actively engaged in research with his professors and colleagues in the areas of online au ctions, healthcare informatics, and computer security. His efforts resulted in five jo urnal articles and eight conference proceedings. As a result of his research efforts, Varol received the University of South Florida Research Excellence award in 2009, and Coll ege of Business Research and Scholarship Award in 2010.