xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim
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-SFE0004744
Comparison of risky decision making processes in dyads and individuals
h [electronic resource] /
by Moumita Mukherjee.
[Tampa, Fla] :
b University of South Florida,
Title from PDF of title page.
Document formatted into pages; contains X pages.
Thesis (MA)--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: The thesis compared the likelihood of taking risks in dyads and individuals in varying situations. Patterns of risky decision making were examined in the standard risky choice task and a novel risk management task. The relative successes of two theories of risky decision making were assessed: Prospect Theory emphasizes perceptual and psychophysical processes, whereas Security-Potential/Aspiration Level Theory emphasizes dispositional and motivational processes. The thesis also examined dyads' decision behavior in light of competing social influence perspectives regarding risky versus cautious shifts and group polarization. Participants, as individuals or as part of a dyad, made decisions in 23 trials about hypothetical two-outcome monetary gambles in one of two different tasks. Risky choice involved making choices between two given 50-50 lotteries which varied in riskiness (i.e., outcome variability), whereas risk management required actively manipulating an existing 50-50 risk by changing outcome values. The 23 trials were equivalent across tasks. Dyad participants communicated via an instant messenger program, while viewing the same lotteries on different computers. Data on risk preferences across gain and loss domains were analyzed using a mixed factorial ANOVA design. Consistent with Prospect Theory value function predictions, the risky choice task led to risk averse preferences for gains and risk seeking preferences for losses, though risk seeking was weak. Consistent with SP/A theory predictions, the risk management task led to overall risk averse preferences, with movement toward risk taking for gains. In addition, there was some evidence of social influences in that dyads tended to be more conservative than individuals in their decision behavior when dealing with undesirable outcomes. Thus, a cautious shift was observed, but only for lotteries involving guaranteed losses. This could not be explained by group polarization. Each of the theories received some support, but none of them could explain all of the findings. Recommendations were made to give greater attention to defining and measuring risk attitudes and dispositions, and to continue exploring differences in decision situations and social settings to obtain a more comprehensive understanding of risky decision making processes. Findings here suggest the need for an overarching theory that can account for a wide variety of influences. A dual processes approach was recommended as one promising avenue. Social and situational influences may prove an essential part of understanding risky decision making in real life contexts.
Advisor: Sandra L. Schneider, Ph.D.
t USF Electronic Theses and Dissertations.
Comparison of Risky Decision Making Pr ocesses in Dyads and Individuals By Moumita Mukherjee A thesis submitted in partial fulfillment of the requirements for the degree of Master of Arts Department of Psychology College of Arts and Sciences University of South Florida Major Professor: Sandra L. Schneider, Ph.D. Michael Coovert, Ph.D. Jennifer Bosson, Ph.D. Date of Approval: October 29, 2010 Keywords: Prospect Theory, Security, Aspi ration, Risky/Cautious Shift, Context Copyright 2010, Moumita Mukherjee
i Table of Contents List of Tables ................................................................................................................ ..... iii List of Figures ............................................................................................................... ..... iv Abstract ...................................................................................................................... ..........v Introduction .................................................................................................................. ........1 Psychophysical Influences on Risky Choices ..........................................................2 Motivational and Dispositional Influences on Risky Choices ..............................11 Social Influences on Decision Making in Groups and Dyads ..............................17 Hypotheses and Predictions ...................................................................................24 Research question 1. Does risky decision making differ across situational contexts? ..................................................................................24 Competing hypothesis 1. ...............................................................24 Competing hypothesis 2. ...............................................................24 Competing hypothesis 3. ...............................................................25 Research question 2. How might risk taking differ from individuals to dyads? ................................................................................27 Competing hypothesis 1. ...............................................................27 Competing hypothesis 2. ...............................................................28 Competing hypothesis 3. ...............................................................29 Research question 3. How might risk taking differ for valence (negative, mixed, and positive lotteries)? .................................................30 Competing hypothesis 1. ...............................................................30 Competing hypothesis 2. ...............................................................31 Method ........................................................................................................................ .......32 Participants .............................................................................................................32 Materials ...............................................................................................................33 Design ....................................................................................................................34 Procedure ...............................................................................................................36 Risky choice task. .....................................................................................38 Risk management task. .............................................................................38 Results ....................................................................................................................... .........41 Discussion .................................................................................................................... ......46 Psychophysical Influences on Risky Decision Making .........................................46 Motivational and Dispositional Influe nces on Risky Decision Making ...............48 Social Influences on Ri sky Decision Making ........................................................49
ii Limitations and Future Directions ........................................................................52 References .................................................................................................................... ......55
iii List of Tables Table 1: Lottery Stimuli broken down by Expected Values and Variability ....................35
iv List of Figures Figure 1: Predictions for Risk Preferences according to Prospect Theory and Risky and Cautious Shift Perspectives ...............................................................28 Figure 2: Predictions for Risk Preferences according to Prospect Theory and Group Polarization Perspectives ........................................................................29 Figure 3: Predictions for Risk Preferences according to SP/A Theory and Risky versus Cautious Shift and Gr oup Polarization Perspectives ..............................30 Figure 4: Sample Screen Shot of a Trial in the Risky Choice Task .................................37 Figure 5: Sample Screen Shot of a Tr ial in the Risk Management Task ..........................39 Figure 6: Effect of Outcome Valence on Risk Preference ................................................42 Figure 7: Effect of Outcome Valen ce and Task on Risk Preference ................................43 Figure 8: Effect of Outcome Valence a nd Decision Maker on Risk Preference ..............44 Figure 9: Effect of Outcome Valence, Task and Decision Maker on Risk Preference ..........................................................................................................45
v Abstract The thesis compared the likelihood of ta king risks in dyads and individuals in varying situations. Patterns of risky decisi on making were examined in the standard risky choice task and a novel risk management task. The relative successes of two theories of risky decision making were as sessed: Prospect Theory emphasizes perceptual and psychophysical processes, whereas Securi ty-Potential/Aspiration Level Theory emphasizes dispositional and motivational processes. The thesis also examined dyadsÂ’ decision behavior in light of competing soci al influence perspectives regarding risky versus cautious shifts and group polarization. Participants, as individuals or as part of a dyad, made decisions in 23 trials about hypothetical two-outcome monetary gambles in one of two different tasks. Risky choice involved making choices between two given 5050 lotteries which varied in riskiness (i.e., outcome variability), wh ereas risk management requir ed actively manipulating an existing 50-50 risk by changing outcome values The 23 trials were equivalent across tasks. Dyad participants communicated via an instant messenger program, while viewing the same lotteries on different computers. Data on risk preferences across gain and loss domains were analyzed using a mixed factorial ANOVA design. Consistent with Prospect Theory value function predictions, th e risky choice task led to risk averse preferences for gains a nd risk seeking preferen ces for losses, though risk seeking was weak. Consistent with SP/ A theory predictions, the risk management task led to overall risk averse preferences, w ith movement toward risk taking for gains.
vi In addition, there was some eviden ce of social influences in th at dyads tended to be more conservative than individuals in their deci sion behavior when d ealing with undesirable outcomes. Thus, a cautious shift was obs erved, but only for lotteries involving guaranteed losses. This could not be explained by group polarization. Each of the theories received some support, but none of them could explain all of the findings. Recommendations were made to give greater attention to defining and measuring risk attitudes and dispositions, and to continue exploring differences in decision situations and social settings to obtain a more co mprehensive understanding of risky decision making processes. Findings here suggest the need for an overarching theory that can account for a wide variety of influences. A dual processes approach was recommended as one promising avenue. Social and situational in fluences may prove an essential part of understanding risky deci sion making in real life contexts.
1 Introduction In real life decision making, individual s often make decisions with another person. Be it a spouse, a parent, a sibling, a friend, a colleague, dyad decision making is common. Much attention has been devoted to studying dyadic interactions and behavior in a variety of contexts. The game theo ry framework provides insight into human cooperative and competitive tendencies in any of a variety of strate gic situations with well-defined rules (e.g., Camerer, 2003). The medical decision making literature addresses issues of shared decision making typically between a patient and his or her physician or health care provider, especially when facing risky alternatives for dealing with serious or life threatening diseases (e.g., Charles, Gafni, & Whelan, 1997; Frosch & Kaplan, 1999; Lgar et al., 2008; Towle & Godolphin, 1999). Th ere is also an extensive literature exploring dyads thr ough the study of intimate relati onships, much of which is relevant to decision making. For instance, st udies often explore how variables such as length and perceived quality of relationship influence relationship-related outcomes, which presumably are the result of decisi ons made within the relationship (e.g., Aron, Norman, Aron, McKenna & Heyman, 2000; Laurenceau, Barrett, & Pietromonaco, 1998). Studies of parents as dyads, or Â“deci sion units,Â” have also been examined, particularly with respect to decisions a bout their children (Becker, 1974; Bostrom, Hoffmann, Krupnick, Adamowicz, Goldman, McWilliams & Varner, 2005). Each of these areas advances our knowl edge of dyad decision making within a specific set of circumstances (e.g., life-threate ning illness) and concerning a particular
2 type of dyad (e.g., father/mother). However, less attention has been given to more generic cases, particularly with respect to day-to-day encounters wi th risky choice. This is somewhat surprising given the quantity of research focused on risky choice by individuals. Hundreds of st udies designed to test hypothe ses derived from expected utility-type theories have produ ced a wealth of data on patter ns of risky choice, typically in the context of two-outcome gambles (see Mellers, Schwar tz, & Cooke, 1998, for a review). In the present st udy, I enlist this classic risky choice paradigm to compare patterns of choice between i ndividuals and dyads. I also move beyond the traditional risky choice paradigm by comparing it to a newly introduced paradigm in which individuals or dyads actively e ngage in management of an ex isting risk. In what follows, I provide a brief overview of studies of individual risky choice, followed by a look at related studies among dyads and small groups. I then introduce some general hypotheses derived from theories dealing with motivation and social processes. Psychophysical Influences on Risky Choices Hundreds of studies have been done expl oring human risky choice behavior (For review, see Levin Schneider & Gaeth 1998; Kuhberger, 1998; Mellers, Schwartz, & Cooke, 1998). Much of the impetus arose from a ttempts to test Expected Utility Theory and later Prospect Theory. BernoulliÂ’s ( 1738, 1954) Expected Utility Theory focused attention on probabilities and outcomes in de scribing risky choices. The theory suggests that subjective values (k nown as utilities) differ from objective values (monetary outcomes) in that the utility of gains increases at a slower rate as values move away from zero. When utilities are graphed as a func tion of objective values, one would expect a straight line if the two were equal to one another. Howeve r, the undervaluing of larger
3 amounts results in a concave curve, or one that exhibits decreasing ma rginal utility. This marginally decreasing utility function predicts that a guaranteed win (e.g., 100% probability of winning $50) would be preferre d over an equal probability of winning $0 or $100, even though the expected values of th e two alternatives are the same. According to the function, the difference between rece iving $0 and $50 would be experienced as larger than the difference between receivi ng $50 and $100, so that it would not be worth giving up the $50 for sure to take a risk on possibly gaining $100. Hence, the concave utility function can explain the commonly obser ved pattern that people tend to be risk averse, avoiding risky alterna tives in favor of safer (high probability) ones. Essentially, BernoulliÂ’s explanation of risk averse behavior relied on the psychophysics of values. Because the experien ced or subjective magnitudes of options systematically differ from the physical magnit udes, risks are routinel y experienced as less valuable than their mathematical expected values. Like other psychophysical principles, this relationship has been confirmed so often that it has come to be known as the law of diminishing marginal utility (e.g., Savage, 1954). Over the years, scholars have suggested variations from the original expected utility theory. The first variation, which dr ew immense attention among economists in the 1940s and 1950s, was introduced by Von Neum ann and Morgenstern (1944, 1947). This version included a canonical mathematical system, via which a rational decision maker would be able to assure coherent and consis tent risky choices, pr ovided he or she was willing to endorse a set of required axio ms. Although von Neumann and Morgenstern did not make any explicit claims about the inhe rent value of the axioms, many scholars entered the debate about the status of thes e decision rules (most not ably, Savage, 1954).
4 Over time, these axioms were widely embraced as requirements for rationality, bestowing the von Neumann and Morgenstern version of expected utility theo ry with a normative status against which actual deci sion making could be evaluated. A second variation to BernoulliÂ’s origin al theory was descriptive in nature. Savage (1954) introduced the concept of subj ective probability in Subjective Expected Utility Theory. SavageÂ’s preferred term was Â‘personal probabilityÂ’ (as originally introduced by Thornton C. Fry), however, w ith time, the term Â‘subjective probabilityÂ’ gained acceptance among scholars. The basic pr emise of subjective probability is that, like values, subjective assessment of probability does not always match the objective probability associated with the outcome in question. SavageÂ’s characterization does not make it clear whether subjectiv e probabilities can best be understood as resulting from psychophysical processes or from individual differences in beliefs about likelihood. The third variation to expected utility theo ry came in the form of Prospect Theory (Kahneman & Tversky, 1979). Like the original expected utility theory, prospect theory starts with the assumption that risky choice behavior is primarily driven by psychophysical processes; in pa rticular, it retains the idea of marginally decreasing sensitivity. Instead of utility, Kahneman and Tversky introdu ce a subjective value function. For gains, the function is concave li ke the standard utility function. However, Kahneman and Tversky also explicitly include losses in their subjective value function. Oddly enough, by maintaining marginally decreasing sensitivity in the negative domain, the function becomes convex, leading to th e somewhat surprising prediction of a tendency to be risk seeking (preferring the gamble over th e sure thing) for losses. Considering a risk of losing $50 for sure versus a 50% risk of losing $0 or $100, the
5 difference between losing $0 and $50 would be experienced as larger (i.e., worse) than the difference between losing $50 and $100, so th at it would not be worth taking a loss of $50 for sure to avoid a possibly larger loss of $100 (marginally decreasi ng sensitivity still at work). This is likely to lead one to go for the risky alternative. Together, this gives the prospect theory value function curve an S-like shape, which is concave in the positive domain and convex in the negative domain. The introduction of a specified origin in the va lue function is also significant. Kahneman and Tversky suggest that perceptual processes ar e highly sensitive to changes and that the subjective value function serves as a measure of change from some default asset position. The origin represents this de fault position and serves as the reference point for choice. This reference point is subject to change with each new deci sion, usually as a function of changes in oneÂ’s status quo. Tver sky and Kahneman (1991) call this reference dependence which implies that the perception of gains and losses is tied to the reference point adopted for the decision. They hypothe size that minor to moderate changes in reference point are not likely to s ubstantially influence preferences. Although Bernoulli briefly touched upon th e issue of losses (Bernoulli, 1954, pp. 26-27), it was the prospect theory value function that clearly showed how choice tendencies are likely to reflect a risk seeking attitude when dealing with perceived losses. Kahneman and Tversky (1979) also argue that perceived losses are experienced more strongly than perceived gains, which is visible in a steeper curve in the negative domain of the value function. They call this loss aversion and argue that, all else equal, Â“losses loom larger than gainsÂ” (Kahneman & Tversky, 1979, pp. 279). For example, people are reluctant to take an even bet that would re sult in either winning or losing some amount,
6 say, $5. Most people would find this bet unattract ive, because the perceived impact of the negative consequence (losing $5) would seem stronger than the potential impact of the equivalent positive consequence (winning $5). Prospect theory also introduces the idea of decision weights. Th e decision weight function retains the basic idea of subjective probability. Kahneman and Tversky (1979) go beyond SavageÂ’s characterization by descri bing decision weights not as simple degrees of belief but as measures of a somewh at more complex construct of Â“the impact of events on the desirability of prosp ectsÂ” (Kahneman & Tversky, 1979, pp. 280). They argue that probabilities are subject to a decision weighti ng function, and the values of possible outcomes are multiplied by these de cision weights. Kahneman and Tversky (1979) theorize that large proba bilities tend to be underweigh ted and small probabilities tend to be overweighted. So people will tend to feel less sure of highly likely outcomes than is warranted and more sure of unlikely outcomes. Prospect theory also incorporates a cogn itive element with the introduction of an Â“editing phaseÂ” in the decision making pro cess. Kahneman and Tversky propose that people usually engage in a Â“preliminary anal ysisÂ” to simplify outcomes and probabilities, prior to entering the eval uation phase. Although Kahneman and Tversky (1979) provide some examples of how editing might occur, th ere is little systematic agreement about these processes. Many studies use simple (Â“ post-editingÂ”) stimuli in order to test the primary predictions of prospect theory a nd rule out editing processes as a possible alternative explanat ion of results. The next section provides a brief review of some of the findings on risky choice. The focus is on the prospect theo ry prediction that is most of ten put to the test (See, e.g.,
7 Hershey & Schoemaker, 1980; Kuhberger, 1998; Levin Schneider & Gaeth 1998; Weber, 1999). The prediction involves a preferen ce reversal pattern in which preferences are risk averse for gains but are risk seeking for losses. Soon after the original prospect theo ry was published, Hershey and Schoemaker (1980) critically eval uated the generali zability of the preferen ce reversal prediction by studying decision making under risk both betwee n and within subjects. They presented participants with a series of choices between a sure gain/loss and a probabilistic gain/loss, with matched expected values. Hershey and Schoemaker found little support for Kahneman and TverskyÂ’s predictions of risk av erse for gains and risk seeking for losses, when making comparisons across subjects. Wh en they did find a reversal (7-25% of choices), it typically involved a sure thi ng with a high value or a gamble with high variance. Given this weak support for Prospect Theory predictions regarding preference reversals, perceptual influe nces on choice behavior may not be sufficient to describe and/or explain ris ky choice behavior. Kuhberger (1998) also completed a revi ew of Prospect TheoryÂ’s preference reversal prediction, but his fo cus was on framing effects and monetary gambles. Framing effects occur when descriptions of the same se t of alternatives in terms of gains versus losses produce preference reversals. A typi cal example of a fra ming manipulation in decision making research is the Asian Dis ease problem. This problem involves choice options framed either in terms of possible liv es saved (positive fram e) or in terms of possible lives lost (negative frame). In hi s review, Kuhberger found a low to moderate impact of framing across studies, resulting in preference differences from positive to negative domains. Kuhberger (1998) also reports preferen ce reversals to be more
8 common in the typical case of riskless choi ce (choices with certain outcomes or sure things) and risky choice (probability of outco mes vary), in comparison to risky-risky choice cases. Thus, while partial support of preference reversals was observed, the question remains how generali zable these patterns are. Levin, Schneider and Gaeth (1998) conducte d a more refined review of framing effect studies. After removing examples of fr aming that did not necessarily involve risky choice, they were able to is olate cases in which risk preferences changed for positively versus negatively framed outcomes. The majo rity of studies showed some effect of framing, though few were clear pr eference reversals as Prospect Theory would predict. In most cases, preferences appeared to diffe r across valence, but not by much. Clear preference reversals were generally found only for studies in which the task domain was similar to the Asian Disease Problem used by Tversky and Kahneman (1981) to introduce framing effects. Several studies did not find significant differences in preferences for gain and loss frames across different scenar ios. This again points out the limited generalizability of Prospect Theory predictions regarding preference reversals from risky gains to losses. Another vein of research focused on differe nt definitions of risk, and showed how different interpretations of ri sk could influence preferences Weber (1999) distinguishes perceived risk attitudes from traditional econom ic definitions of risk as variance. For economists, risk aversion is associated with low variance outcomes a nd risk seeking with high variance outcomes. Perceived risk attit ude emphasizes the indi vidualÂ’s perception of some alternative as more or less risky. Weber and colleagues f ound that the average individualÂ’s perception of risk may not be exclusively dependent on variance. The same
9 alternative (for example, the low variance outco me alternative) could be perceived by the decision maker as less risky in one doma in and more risky in the other. Using gamble pairs, Mellers, Schwar tz, and Weber (1997) found economicallydefined risk attitudes to be re versed from gain to loss domains for the majority of subjects (61%), which is consistent with Prospect Theory predictions (though somewhat weak). However, they observed that perceived risk attitudes did not change in the same way. Most participants (60%) were consistent acro ss domains in their perceived risk attitude, with 44% of all participants reporting that th ey were risk averters. This points out that traditionally defined concepts of risk may not always fit with what people experience as risk. These findings also raise a question abou t Prospect TheoryÂ’s ab ility to explain risky choice in terms of perceptual or psychophys ical processes. Logically, the Prospect Theory implications regarding willingness to take risks and the perc eived risk attitudes studied by Weber and colleagues should matc h up. However, participantsÂ’ responses indicated that their (and pe rhaps the lay personÂ’s) inte rpretation of risk does not necessarily equal the ec onomic definition of risk assumed in Prospect Theory. This casts some doubt on the classification of Prospect Theory as a perceptual theory of risky choice. Weber and Milliman (1997) studied perceived risk attitude and ri sk preference in paradigms of commuter train times and financial investments (with hypothetically endowed amounts). They also found that participan ts tended to be cons istent in their risk preferences based on their assessment of percei ved riskiness of altern atives. Over 75% of the participants exhibited c onsistent perceived-risk at titudes across gain and loss domains, and more than 65% consistently ch ose the less risky alternative across domains
10 in both the commuter train times and financ ial investments studies. They observed risk seeking preferences in the gain domain, when participants expressed their preferences for commuter train times. Thus, results were not consistent with preference reversal predictions of Prospect Theory. Additionall y, here too, the concept of perceived risk attitude, as reported by participants, di d not match with the standard economic interpretation of risk as variance. Weber and colleaguesÂ’ work provides weak support for the kinds of preference reversals predicted by Prospect Theory, especi ally in cases involving sure things and high risk situations. At the same time, their work also highlights problems with the basis for Prospect Theory predictions. Prospect Th eory may not adequately capture peopleÂ’s experience of riskiness, thus the perceptu al underpinnings of the theory may be in question. In summary, there is mixed support for Prospect Theory preference reversal predictions. The prediction is most likely to hold for c hoices involving a risky option compared to a sure thing. The lack of ge neralizability points out that there may be additional influencing factors on risky choice than those prop osed in Prospect Theory. Most people tend to have a strong reaction when they are exposed to risk, which may be reflected in their perceived attitude toward ri sk. As we have seen, this experience is not reflected in the economic view of risk. One as pect of perceived risk that is ignored in both economic theory and Prospect Theory vi ews of risky choice is an affective or motivational component which intuitively is an integral part of ri sky situations. People routinely describe the experi ence of hope and fear associat ed with taking risks. In addition, pre-existing orientations or personality characteristic s, may also play a role in
11 determining how people react to risk in genera l. Therefore, looking into dispositional and motivational factors may give us another pers pective on peopleÂ’s dea lings with risk. The next section reviews briefly appro aches exploring this possibility. Motivational and Dispositional Influences on Risky Choices Another approach to risky decision making explains the process as arising from dispositional and motivational sources, rather than psychophysical ones. Lopes introduced the SP/A (Security Potential/Aspira tion) theory, which brings in affective influences that are more in line with tr aditional approach-avoidance paradigms (Lopes, 1984, 1987; Schneider & Lopes, 1986). Lopes does not deny the relevance of the marginally decreasing utility function. She rath er draws attention to the differences in the theoretical approaches to describing decision behavior. Lopes (1984, 1987) suggests that there are two dispositional inc linations towards risk that differ according to appetitive versus avoidant reactions to risk. Those with a security focus are dispositionally inclined to avoid the negative consequences associat ed with risk taking and those with a potential focus are dispositionally inclined toward approaching the positive consequences associated with risk taking. Lopes suggests that most people are se curity oriented, and are thus typically risk av erse in their behavior. SP/A theory also suggests that situati onal factors introduce another approachavoidance variable, in particular, with respec t to aspiration levels. High aspiration levels typically require that some risk must be tole rated in order to reach a goal, whereas lower aspiration levels may be reached without having to take risks. SP/A theory implies that risk preferences will be str ong when dispositional tendencies match situational needs, but
12 will be weaker and more conflicted when dispositional tendencies are at odds with situational needs. According to Lopes, most people are likely to exhibit risk aver se decision patterns most of the time, because most people are hypothesized to have security-prone dispositions. There are two occasions, howev er, when even security-minded people may engage in risk-seeking behavior. First, when they feel safe (i.e., when no element of threat to their sense of security exists), they may feel comfortable taking risk. So, for instance, security-minded people might be wi lling to take a risk when all outcomes involve something positive. The second occasion is on the other extreme, when securityminded people are under great thr eat. In such threatening situ ations, no safe alternative is available, and any hope of getting out of the si tuation may require taking a risk. In these cases, the disposition to Â“play it safeÂ” is in conflict with the aspiration to avoid acceptance of a bad outcome. Instead of using two-outcome gambles, Schneider and Lopes (1986) used a more complex set of multi-outcome lotteries to st udy decision preferences of individuals who were pre-screened as dispositionally risk aver se or risk seeking. They found 70% of their prescreened participants to be dispositionally risk averse (risk averse participants). Schneider and Lopes (1986) observed comple x patterns of risk preferences, which emphasized that, factors other than percepti on must also be at work. They found weak support for reversal of preferences from risk averse to risk seeking in gain to loss domains predicted by Kahneman and Tversky (1979). With the exception of lotteries with a better-than-zero (risk less lottery) worst outcome, the typical risk averse participant was strongly risk averse for ga ins. Hence, the sure thing was strongly
13 preferred in most cases. Lo tteries that guaranteed more than zero were also quite popular, and sometimes preferred to the sure thing. Lotteries with minimum outcomes greater than zero pose no threat to the sense of security, since there is nothing to lose with these gambles. This may have allowed the pa rticipants to raise their aspiration level higher than the sure thing. The riskless lotteries offer the opportunity for winning an amount higher than the sure amount, possibly cl oser to the raised aspiration level. However, in the loss domain, the possibility of losing at least some amount posed a serious threat to the participantÂ’s sense of safety. Therefore, for loss lotteries, the typically risk averse participant exhibited a mixed pattern of pref erences. For losses, participants appeared to dislike the lotteries at both extremes; the safest (including sure thing) probably because it does not meet their aspiration levels, and the riskiest, probably because the threat to security is too high, a nd they wanted to minimize the chance of the worst loss. HigginsÂ’ Self Discrepancy Theory (1 987) and later the Regulatory Focus Theory/Principle (1997) also share a focu s on motivational and di spositional factors influencing human experience and behavior. HigginsÂ’ earl y work centered on explaining personality differences in terms of disposit ional tendencies to focu s on discrepancies in self-concept relative to an Â“ide alÂ” self or relative to an Â“ oughtÂ” self. Higgins found that those who focus on self-concept discrepancies with the Â“idealÂ” self are more apt to regularly experience emotions such as joy and dejection, whereas those who focus on discrepancies with the Â“oughtÂ” self are more likely to expe rience feelings of relief and anxiety. These findings eventually led to HigginsÂ’ concept of regulatory focus, which highlights motivational influences in personalit y. In particular, Higgi ns hypothesizes that
14 some people tend to focus on promoting the pos itive, whereas others prioritize preventing the negative (Higgins, 1998; Shah & Higgins, 1997). Higgins connects an individualÂ’s ideal self discrepancy concerns with promo tion focused strategic in clinations, where one seeks matches with positive outcomes, such as hopes and aspirations. Higgins connects the individualÂ’s ought self discrepancy c oncerns with prevention focused strategic inclinations, where one seeks prevention of mismatches with ne gative outcomes often associated with duties and obligations, (Higgins, 1996; Higgins, Roney, Crowe, & Hymes, 1994). The prevention-promotion distin ction in regulatory focus is conceptually similar to LopesÂ’s security-potential di chotomy. Both emphasize the importance of approach-avoidance motivation as a guide in behavior. Crowe and Higgins (1997) studied the im pact of regulatory focus on strategic inclinations in decision making, using a si gnal detection paradigm. Crowe and Higgins (1997) associated a risky response bias w ith promotion-focused individuals and a conservative response bias with preventio n-focused individuals. They found that promotion-focused individuals we re more likely to take risk s (be more prone to getting Â‘hitsÂ’ and Â‘false alarmsÂ’), and prevention-focu sed individuals were more likely to play it safe, and generally be risk averse (be more prone to avoiding Â‘missesÂ’ and getting Â‘correct rejectionsÂ’). This finding may be mapped on to Lopes (and colleaguesÂ’) work connecting potential-oriented pe rsons with risk seeking and security-oriente d people with risk aversion (Lopes, 1984, 1987; Schneider & Lopes, 1986). Weber and Milliman (1997) also drew th e connection between dispositional and situational factors in their explanation for risk preferences and perceive d risk attitudes. In choices between possible commuter train times they observed risk seeking (i.e., higher
15 variance) preferences for 61% of the participan ts in the gain domain. This does not match what would be predicted by the Prospect Th eory value function. Weber and Milliman (1997) cite aspiration level as the possible influencing fact or behind peopleÂ’s preferences for the higher variance train times in the gain domain. When all options were faster than average, people could rely on some savings e ither way, so that none of the options would have a downside orÂ—in SP/A termsÂ—be a threat to security or the status quo. Thus, aspiration levels could be rais ed without any worry, so that the possibility of saving the maximum time could now be considered the more attractive alternative. Risk perceptions were also apparently sensitive to this lack of a downside. Greater unpredictability, when there was no downside, was associated with th e possibility of the gr eatest savings, and so was seen by some (34%) as less risky. In the loss domain, where trains were running slower than current average commute time, however, concern for avoi ding the maximum time loss frequently led participants to make risk averse choices in traditional terms. In this case, greater unpredictability had a downside, or an elemen t of threat (e.g., being late for work or class), which was seen by the overwhelming ma jority as more dangerous or risky. Weber and Milliman (1997), therefore, emphasized that what people considered as Â“riskyÂ” changed from one domain to the other. These observations also point out that real life contingencies may be a determining factor behind assessments of riskiness. Dispositional and motivationa l theories have thus informed researchers of the possibility of pre-existing tende ncies and situational character istics together influencing peopleÂ’s behavior. They remind us that peopleÂ’ s initial (gut-level) reaction towards risk
16 may set the tone for the decision making pro cess; but the complete process may involve impact of situational demands as well. From a review of the existing litera ture on risky decision making, Prospect Theory has done fairly well in describing st andard risky choice be havior, particularly when sure things or extreme outcomes have been involved. Nevertheless, it is not clear that Prospect TheoryÂ’s reliance on psychophysics gets to the heart of the experience of risk, and thus, may be inadequate or incomplete as a theory of risky choice. Motivational and dispositional explanations, including the SP/A Theory, potentially may expand the ability to understand risky choice across a larg er variety of risky decision situations. With the objective of exploring these di fferent influences on risky decision making, the present study includes two different tasks. I hypothesize that the simpler psychophysical explanation is more likely to be plausible in situa tions that are more superficial and less actively engaging. In contrast, motivational and dispositional influences are more likely to be seen when the decision situation demands a more active level of engagement on the part of the decisi on maker. The first task is the standard passive risky choice task, which is a para digm expected to favor the perceptual perspective. The second task is a novel activ e managing risk task (Schneider, Hudspeth & Decker, manuscript in preparation), which ma y be more likely to engage dispositional and motivational processes. Additionally, the author explores deci sion making in dyads, which provides the opportunity to examine whethe r the introduction of anothe r person into the decision making environment may bring in yet another se t of factors: social influences. Findings related to possible social influences on risky choice will be discussed next.
17 Social Influences on Decision Making in Groups and Dyads While group phenomena have been studied extensively in vari ous sub-areas of psychology throughout the last century, studies of decisi on making under conditions of risk comparing groups and individuals became popular starting in the early sixties. Hunt and Rowe (1960), Stoner (1961) and Wallach, Kogan and Be m (1962) are a few names among researchers who studied risky decision making in individuals and groups in the early sixties. Wallach et al., (1962) found groups to be more inclined to prefer risky alternatives over cautious one s following group discussions, as compared to individuals in the absence of discussion. In Wallach et al.Â’s, (1962) study group members were previously unacquainted and had the same status (no identifiable influential characteristics) when group members began in teracting. Participants considered several scenarios, and in each they provided th eir recommendation regarding the lowest probability of success that the ch aracter in the scenario should require before selecting the riskier of two options. In questions afterward, some individual members of the group reported that individuals who we re more risky tended to be more influential in the group than those who were more cautious. Wallach et al., (1962) offer two possible e xplanations for the observed risky shift phenomenon. First, Wallach et al., (1962) cons ider Â‘diffusion of responsibilityÂ’ among group members as a possible reason for the shift towards riskier decisions following group discussion. Diffusion of responsibility o ccurs when being part of a larger group reduces individual memberÂ’s sense of persona l accountability or res ponsibility for the outcomes of decisions. Wallach, Kogan and Bem (1964) speculate that this phenomenon
18 may be related to decreased anxiety from emotional bonds created in the group setting, where risk is perceived as shared. Second, Wa llach et al., (1962) consider the possibility that risk takers tend to be more influential in groups and that this could play a role in the observed shift towards riskier decisions following group discussion. They propose that these individuals might be more inclined to take initiative in mediating the group discussion. Such initiative could lead to grea ter acceptance of risk in the decision making, compared to average pre-discussion individual preferences. With respect to the present thesis, the que stion is whether we shall observe similar decision patterns in dyads. Previous work s uggests that the presence of many people in a group provides the possibility that a given individualÂ’s sense of accountability may be diminished. However, when only two people ar e present, as in a dyad, both are at least partly responsible for the decision outcomes Therefore, for dyads, it is questionable whether diffusion of responsibility would o ccur. Also, with two people, the hypothesis that risk takers might be more influentia l seems less likely given that both people are required to interact to make the decision. Using a similar design as the one used by Wallach et al., ( 1962), Stoner (1961) compared risky decisions in groups and individu als. He observed a risky shift in majority of his participants. However, not all group me mbers were consistently more risky across all items following discussion. Nordhy (1962) revisited Stoner ( 1961) and Wallach et al.Â’s (1962) work. He observed that, for at least some of their choice items, several group members exhibited a cautious shif t following group discussion. Stoner (1961, 1968) explains this with re ference to the general value hypotheses of Nordhy (1962) and Brown (1965). According to this hypothesis, group decision shifts
19 would be to riskier or safer (more cautious ) alternatives depending on how they match with widely held values within the culture, which will be reflected in the individualÂ’s original inclinations. For example, if deci sions concern a choice such as marriage, the culturally approved value may rest on the side of caution rather than risk; in such cases, group discussion is likely to l ead to more cautious decisions compared to individuals making decisions by themselves. The risky and cautious shift observations we re later described as variations of a single phenomenon called group (attitude) pola rization (Moscovici & Zavalloni, 1969). A wide range of theoretical and research work has been carried out in the area of group polarization, from the pioneer ing work of Moscovici and Zavalloni (1969), to Myers and Lamm (1976) to more recent work by Stasse r and Titus (2006), and Luhan, Kocher and Sutter (2009). Group polariza tion occurs when discussion among the group members creates the sense of a single decision making unit, working towards a shared goal. This often causes individual preferences to align wi th that of the group ma jority. Inclination of the majority of the individual members in a group is strengthened when the group behaves as a decision making unit. Research with groups has shown that groups make significantly Â“more extremeÂ” choices than in dividuals (For review see Isenberg, 1986; Myers & Lamm, 1976). Notably, decisions tend to be made in the same directions as majority of the individuals. Research has shown that i ndividuals who made moderate ly risky decisions prior to group interaction made even riskier c hoices following group discussion, whereas individuals who made initially cautious pref erences showed a hei ghtened tendency to choose safer alternatives following gr oup discussion (e.g., Deets & Hoyt, 1970;
20 McCauley, Stitt, Woods & Lipton, 1974; Zaleska, 1974, 1976). Thus, the polarization effect produces the same pattern of pre-di scussion preferences, only more pronounced. Notice that the risky or cautious shift is some times consistent with polarization processes, but not always. If group discussion shifts decisions in a direc tion opposite to the one originally favored by many individuals (e.g., risk y shift in which risk averse individuals gravitate to more risky choi ces after discussion), this w ould be inconsistent with a polarization explanation. For the present thesis, the relevant question is concerned with how dyadsÂ’ decision preferences may be affected by gr oupÂ—or in this case, pa rtner-discussion. If polarization takes place in dyads, dyad memb ers will make more extreme decisions together compared to individuals. So, if indi viduals are typically ri sk averse, dyads would be expected to be even more risk averse. If individuals are risk seeking, dyads will be even more so. If risky or cautious shifts occur, dyads will tend to take more of less risks, respectively, than individuals, indepe ndent of individual predispositions. There are few studies that allow a compar ison of risky/cautious shift versus group polarization hypotheses in the c ontext of dyads. In one of the rare exceptions, Keller, Sarin and Sounderpandian (2002, 2007) compared individual and dyad preferences for monetary gambles involving risk (i.e., stated probabilities) or uncertainty (i.e., unknown probabilities). In all conditi ons, they found that cautious shifts predominated. Both individual and dyad tended to be risk aver se. Most dyads (typically around 60%) became more risk averse (i.e., cautious) than i ndividuals when facing risky and ambiguous monetary choice situations. However, there was also a sizable minority that became less risk averse (i.e., risky shift, typically around 25-30%). Alth ough the majority results are
21 consistent with a group polarization explanat ion, the subset of dyads that moved in a riskier direction suggest that other so cial influences we re also present. In a more recent study, Hardisty and Sanitioso (2007) studied three-person groups using various communication media to make hypothetical decisions about taking business risks. Hardisty and Sanitioso found that face-to-face discussion as well as video conferencing led to more extreme group deci sions compared to individual ones. The pattern of results was consiste nt with Prospect Theory pred ictions of risk aversion for gains and risk seeking for losses, though more extreme for groups than individuals. However, they did not find this effect of polarization for instant messaging (IM). For the IM condition, participants showed an opposite effect, expressing risk aversion for losses, and moving toward risk seeking for gains (t hough in both conditions, they were generally in favor of the sure option). Again, ther e is partial support fo r the role of group polarization but only for tw o of the three media. Work on peer influences may also provide some insight into social influences in dyad decision making. Gardner and Steinbe rg (2005) studied the impact of peer influence on risk preference, risk ta king, and risky decision making by comparing individuals and three person gr oups in adolescents (13-16 year olds), youths (18-22 year olds), and adults (24 years and older). The authors tested risk-tak ing using a video game about driving called Â‘Chicken.Â’ In the video game, participants were required to make decisions about stopping or continuing to driv e after a traffic light turns yellow. While continuing could earn more points for the partic ipant, crashing if the light turned red before the car stopped would lead to losing a ll points. Gardner and Steinberg found in all three age groups that individuals engaged in more risk taking when in presence of two
22 other same-age individuals volunteering ad vice than when alone. They observed the overall impact of peer influence (enhancing ri sky inclinations) to be most prominent in adolescents and young adults compared to olde r adults. The individua ls in peer groups thus exhibited a risky shift, especially among adolescents. Because our dyad participants are mostly in the youth (young adult) age range, we are likely to observe some effect of peer influence on risk preferences. Alt hough it was not possible in Gardner and SteinbergÂ’s study to assess wh ether the shift in preferen ces was consistent with polarization of risk attitudes, the present work will provide the opportunity to observe whether polarization versus risky shift patterns occur in dyads. In addition to peer influences, the li terature on persuasion may also provide insight into social influen ces in decision making. WoodÂ’s (2000) review of the persuasion and social influences in social interactions revealed that motivations about the self, the other and the situation may a ll bring about changes in atti tude. One way that social influence may occur is through le arning new arguments in favor or against an alternative. According to the classic Persuasive Ar guments Theory (Burnstein & Vinokur, 1977; Vinokur & Burnstein, 1974, 1978; Vinokur, Trope & Burnstein, 1975), for instance, each member of a group is only aware of a few of the existing arguments in favor of a choice. Group interaction offers the opportunity for exposure to a greater number of arguments compared to deciding alone. If discussion leads to a greater number of arguments favoring riskier over safer alternatives, the Persuasive Arguments Theory would suggest a riskier choice, and vice vers a (Vinokur, Trope & Burnstein, 1975). In the context of the present study, dyads may potentially come up w ith more arguments in favor of the riskier (or safer) alternatives and therefore make decisions towards (or away from) risk
23 depending on cognitive availabi lity of arguments. Accordin g to SP/A theory, since most people are likely to be more concerned with s ecurity, they may also be more likely to think up arguments in support of caution rather than risk, leading to an increase in risk averse decisions. Finally, one of the other fact ors that could be a possibl e social influence on the dyad decision process is nor ms. According to Fehr and Fischbacher (2004), from food sharing to mating practices to religious trad itions, and cooperative as well as defense activities, social norm s have a large role in influenc ing human behavior. Clark (1984), Elster (1989), Fiske (1992) and Heyman and Ariely (2004) observed the application of different sets of norms by people when dealing with different sets of scenarios. Out of these, two major categories have been noted: social and economic (market) norms. Compared to the economic market, which usually involves a clear exch ange of equivalent benefits (e.g., a person expect s to pay a certain amount of money to receive a good of a given value), the social context is more co mplex. The social context involves different types of interactions, and different types of norm relations (e.g., a person may not expect to pay for a good given to them by a pa rent). Additionally, there are varying circumstances, in which the same relations can take on different considerations (e.g., a friend becomes oneÂ’s work supervisor), requ iring reassessment of norms to be applied. Therefore, in social interactions, the norms may be harder to determine and apply a priori. Kerr, Garst, Lewandowski and Harris (1997) studied the influence of human tendency to behave according to established gr oup norms compared to inclination to act following internalized personal norms. They found that persona l convictions were not the only predictor of how people behaved, but that group norms, which were defined and
24 established via group discussion, also contributed to behavior In working with dyads, we might also expect that social norms may become a part of the decision process. Because the dyads will be working c ooperatively toward a common ou tcome, we expect that social norms would be likely to enhance collaborative efforts. In the context of the present research, we imagine that dyad partners will be motivated both to persuade their task partne r towards their own way of thinking, but also to work cooperatively toward a common goal. If security is a more common motivation as predicted by SP/A Theory, the dyads may be able to accomplish both goals by tending toward risk averse choices. Th e next section describes in de tails the general and specific hypotheses. Hypotheses and Predictions The thesis presents the opportunity to explore whether individuals and dyads respond similarly when dealing with risk in different situations. For each research question, competing hypotheses supported by each relevant theoretical perspective are delineated. Research question 1. Does risky decisi on making differ across situational contexts? Competing hypothesis 1. If Prospect Theory value function predictions hold across multiple situations, then for both risky choice and risk management tasks, participants would exhib it a risk averse pattern of prefer ences for gains and a risk seeking pattern for losses. Competing hypothesis 2. If SP/A theory predictions hold across multiple situations, then for both ris ky choice and risk management tasks, participants would
25 exhibit a general risk averse pattern of preferences. Some risk taking may be observed, particularly for choices invol ving assured gains, and posing no threat to security. Competing hypothesis 3 If risky decision making differs due to the nature of the situation, then the two theories may appl y differentially based on the psychological processes most likely to be activated. The Risky Choice task seems more passive, static, and reactive whereas Risk Management was designed to be more active, dynamic, and situated. Therefore prospect th eory may be better able to pr edict risky choice preferences whereas SP/A theory may be better able to predict risk management preferences. Prospect Theory is a theory of choice (with the characteristics listed above); so Prospect Theory should do a good job of pred icting results for the choice task, risk seeking for losses and risk averse for gains. Since stimuli are in e ffect equivalent across tasks, Prospect Theory should do just as well at predicting the management task unless the process of actively changi ng oneÂ’s current situation is fundamentally different from passively responding to given alternatives Active decision making may engage motivational processes more, which could change the pattern of decision making. SP/A theory is a theory about how disposit ions and situational motivations affect risk preferences. Ideally, the theory should be applicable to both choice and risk management tasks. However, the decision ma king literature shows risky choice behavior to be fairly consistent with prospect theo ry predictions, but thes e are all for standard (passive) risky choice situations and usually situations invo lving a sure thing versus a two-outcome risk where one outcome is zer o. SP/A may do better for other gain and loss lotteries (ones that may not involve a sure thing or zero outcome).
26 Risk management, by definition, is a mo re active, dynamic undertaking that is likely to initiate motivational processes that are more typical of daily-life decision making. In particular, because management i nvolves dealing with a situation over time, it might better reflect the combin ation of dispositional, mo tivational, and situational influences hypothesized in SP/A theory. Tende ncies toward ensuring security or seeking potential might be exaggerated due to th e active manipulation of worse and better outcomes in risk management. Aspiration levels might be more salient in the management task than the choice task because pa rticipants may set a level as part of their management strategy. Thus, overall, SP/A theory predictions may be more predictive in active risk management tasks compar ed to passive risky choice tasks. Schneider and colleagues (Schneider, Hudspeth and Decker, manuscript in preparation) showed that, in individuals, tasks focused on newly introduced risk (as in standard risky choice) yielded remarkably di fferent patterns from an analogous situation that was presented as a problem of managi ng existing risk. In th e first (risky choice) study, they found a weak pattern of results rough ly consistent with prospect theory with risk seeking in negative and risk aversion in the positive domain. However, in the risk management study, they observed a distinctly risk averse approach on behalf of the individuals when approaching ri sk overall. In a sense, in the risk management study, the individual (or the dyad in the present study) is endowed with this (e xisting) risk and has to manage it. This active aspect of risk management is likely to be influenced by motivational factors, more so perhaps than the passive risky choice task. In the present work, the opportunity was avai lable to extend these findings to decision making in dyads.
27 Research question 2. How might risk taking differ from individuals to dyads? The social psychological literature offers at least two relevant perspectives regarding influences of social factors on risky decision making. According to the risky/cautious shift perspectiv e, groups (in the present wo rk, dyads) would engage in greater/lesser risk taking follo wing group discussion. This shif t would be irrespective of the original attitude of majo rity of the individual partic ipants. According to the group polarization perspective, on th e other hand, dyads would make more extreme decisions in the same direction as the majority of i ndividuals. Because both SP/A and Prospect Theory posit different risk preference patte rns as a function of outcome valence, the prediction for group polarization assumes that individualsÂ’ attitudes toward risk may vary for gains and losses. Therefore, preferences of individuals at various outcome levels were used as the indicator of majority indi vidual attitudes for ga ins versus losses. Competing hypothesis 1. According to the risky shift literature of group decision making, groups tend to take more risks than individuals. In contrast, the cautious hypothesis suggests that groups avoid risk more than individuals. If risky or cautious shift and Prospect Theory (PT) were both to be correct in predicting individualsÂ’ and dyadsÂ’ risk preferences, results would show a pattern illustrated in Figure 1. So, for PT and risky shift predictions to be correct, dyads would tend in all cases to make riskier decisions comp ared to individuals as shown in the dotted line above the individual data. On the other hand, if PT and cautious shift predictions were correct, dyads would consistently make safer choices than individuals as indicated by the more risk averse preferences in the dotted line below the individual data.
28 Figure 1. Predictions for Risk Preference s according to Prospect Theory and Risky and Cautious Shift Perspectives Competing hypothesis 2. Assuming that individual majority preferences are indicative of typical risk at titudes, the group polarization hypothesis would suggest that dyad discussion would strengthen those initial at titudes and exaggerate patterns of risk behavior. However, this shift could be either in the direction of risk or caution, depending on dyad partnersÂ’ original predispositions. If Prospect Theory and group polarization perspectives were both correctly predicting ri sk preferences, deci sion behavior patterns would be likely to resemble the dotte d preference curve in Figure 2. Overall, dyads would continue the same pa ttern of preferences of risk seeking for losses and risk aversion for gains, as predicte d for individuals in PT. If group polarization took place, the only notable change would be an emphasized pattern for the dyads in each domain. Therefore, dyads would tend to be more risk seeking for losses and more risk averse for gains compared to individuals. Predicted Risk PreferenceOutcome Valence Losses -GainsProspect Theory & Risky/Cautious Shifts Predictions Individuals Dyad Risky Shift Dyad Cautious Shift Risk Seeking Risk Aversion $0 0.50
29 Figure 2. Predictions for Risk Preference s according to Prospect Theory and Group Polarization Perspectives Competing hypothesis 3. If risky or cautious shift and SP/A Theory were both to be correct in predicting indi vidualsÂ’ and dyadsÂ’ risk prefer ences, results would show a pattern illustrated in Figure 3. For SP/A and risky shift pred ictions to be correct, dyads would make riskier decisions compared to in dividuals as shown in the dotted line above the individual data. On the other hand, if SP/A and cautious shift predictions were correct, dyads would consistently make more risk averse choices than individuals as indicated by the dotted line below the individual data. If group polar ization were coupled with SP/A theory predictions, dyad preferences would resemble the same pattern as for cautious shift because majority preferences under SP/A theory are expected to be predominantly risk averse across most if not all outcome values. Predicted Risk PreferenceOutcome Valence Losses -GainsProspect Theory & Group Polarization Predictions Individuals Dyad -Group Polarization Risk Seeking RiskAversion 0.50 $0
30 Figure 3. Predictions for Risk Preferen ces according to SP/A Theory and Risky versus Cautious Shift and Group Polarization Perspectives Few studies have explored group versus individual preferences across outcome valences. One exception, Marquis and Reitz (1969) observed that groups would take more risks than individuals for positive expect ed values, but were less likely to take risks for negative expected values, with no discerna ble change for zero expected value. This was consistent with the original individua l preference patterns suggesting that group polarization processes provided a better explanation than either a general ris ky shift or a cautious shift. These results also provide s upport for the idea that social influences may affect risk taking differently as a function of outcome valence (a nd in this case were generally supportive of the pa ttern anticipated in SP/A theory rather than PT). Research question 3: How might risk taking differ for valence (negative, mixed, and positive lotteries)? Competing hypothesis 1. According to Prospect Theory, for positive lotteries which would result in gains, risk averse preferences are expected. For negative lotteries Predicted Risk PreferenceOutcome Valence Losses -Gains SP/A, Risky/Cautious Shift & Group Polarization Predictions Individuals Dyad Risky Shift Dyad Cautious Shift + Group Polarization RiskSeeking Risk Aversion$0 0.50
31 representing losses, a genera l inclination towards risk ta king would be observed. For mixed lotteries (wherein outcomes include losse s, gains, and/or zero), preference patterns are likely to be exaggerated because the value function implies that sensitivity to changing values should be greatest for values closest to zero. Fo r negative expected values, an intensified risk seeking pattern of preference s would be expected, and for positive expected values, a stronger risk averse pattern would be expected. Competing hypothesis 2. According to SP/A theory, an overall risk averse pattern is predicted. Loss lotteries present a more ta ngible threat to peopleÂ’ s sense of security, supporting risk aversion. For gain lotteries, no tangible threat to security is present. Therefore, if aspiration level is set (modera tely) high, the individual may engage in some risk taking. For mixed lotteries, concern for security would push decision makers away from outcomes with losses. For lotteries wi th negative expected values, this would typically lead to risk aversion in order to avoid the worst possible loss, but occasionally might support risk seeking if th e riskier option is th e only one that invol ves the chance to lose nothing. For lotteries with positive exp ected values, security-seeking would lead to strong risk aversion when the riskier option involves a possible loss, and weaker risk aversion when the riskier optionÂ’s worst outcome is zero.
32 Method This thesis brings together findings from four different stud ies, comparing risky decision making in dyads and individuals in two different tasks. One task is the standard passive risky choice task, which requires participants to react to pre-determined risky alternatives. The other is a novel managing ri sk task, which require s participants to actively modify a risk that th ey are currently facing. Only data for dyads were collected as part of the thesis. Data for individual s were collected by Schneider, Hudspeth and Decker (manuscript in preparation) earlier. E ach set of data was collected in a different semester, from virtually the same particip ant pool of undergraduate students enrolled in Psychology courses, with a high consistenc y of demographic characteristics across semesters. Participants Data were collected following university IRB requirements from students enrolled in undergraduate Psychology courses for course credit. After an introduction to the lotteries, a quiz was administered to ensure that participants u nderstood the task. The quiz assessed basic understandi ng of probabilities and how probabilities relate to outcomes in lotteries. It also tested whethe r participants comprehended the computerized representations used to disp lay the lotteries. Data for participants who failed the quiz were not included in analyses. For dyads, bot h participants had to pass the quiz for their combined data to be included in analysis.
33 For individuals, 47 out of 63 participants passed the quiz in the risky choice task and 69 out of 79 participants passed the qui z in the risk management task. For dyads, both members passed the quiz in 26 out of 44 dyads in the Risky Choice task and 42 out of 56 dyads in the Risk Management task. Al though there was considerable variation in pass rates, this was not unusual given extensiv e data on semester-by-semester variability in similar studies. Materials The same twenty-three pairs of monetary lo tteries were used (i n different ways) as stimuli in each of the risky decision making tasks. Each lott ery in a pair could result in one of two equally probable outcomes, i.e., ev ery lottery was composed of two outcomes, each with a 50% chance of occurri ng. One lottery in each pair was riskier than the other. The riskier lottery had tickets that were farthe r apart in value, representing more extreme potential outcomes. The less ri sky lottery had tickets that we re more similar in outcome value, offering the participants a better f unctional sense of where they might end up. Table 1 shows the complete set of lottery pairs that were used. The pairs of lotteries covered a wide range of outcome values from -$300 to +$300, and were classified in terms of outcome valences as all nega tive (NN; loss lottery pairs), mixed negative (MN), mixed positive (MP) and all positive (PP; gain lottery pairs). Each outcome valence condition was ma de up of six lotteries that were factorially determined using a 3 x 2 Variance x Relativ e Expected Value (RelEV) design. Variance conditions were defined according to three de grees of ticket separa tion. For the high variance lottery pairs, the out comes of the riskier lottery were separated by $200 and the outcomes of the less risky lottery were se parated by $100. For the medium variance
34 lottery pairs, the outcomes in the riskier a nd less risky lotteries were separated by $150 and $50 respectively. For the low variance lottery pairs, the ou tcomes in the riskier and less risky lotteries were sepa rated by $100 and $0 (zero) resp ectively. Relative expected value was defined by relative outcome distan ce from zero. The outcome that is farther from zero would count as high relative EV while the outcome closer to zero would count as low relative EV. The 3x2 factorial construction of the lo ttery pairs in each outcome valence set was adopted to systematically ensure a comprehensive range of stimuli. Although Variance and Relative Expected Value have been analyzed elsewhere (Schneider, Hudspeth & Decker, manuscript in prepar ation), a detailed exploration of these influences on decision making are beyond the scope of the thesis. Design The study used a 2 x 2 x 4 (Decision Ma ker [individual, dyad] x Task [risky choice, risk management] x Outcome Valen ce [all negative, mixed negative, mixed positive, all positive]) Repeated Measures Factorial design. Decision Maker was a between-subjects variable representing whet her the decisions were being made by an individual or a dyad The data for dyads were collected as part of the MasterÂ’s thesis, and compared to existing data for individuals (w hich were collected pr eviously by Schneider, Hudspeth, & Decker, manuscript in prepara tion). Task was also a between-subjects variable; some of the subjects completed the Risky Choice task and the remainder completed the Risk Management task. Outcome Valence was a within-subjects variable with 4 levels: all negative (NN, losses), mixed negative (MN), mixed positive (MP) and
35 Table 1. Lottery Stimuli broken down by Expected Values and Variability Risk Seeking Risk Averse Lottery (by Trial) Valence Expected Values Rel EV Variance Low Ticket High Ticket Low Ticket High Ticket B NN -200 High High -300 -100 -250 -150 C NN -175 High Medium -250 -100 -200 -150 D NN -150 High Low -200 -100 -150 -150 E NN -150 Low High -250 -50 -200 -100 F NN -125 Low Medium -200 -50 -150 -100 G NN -100 Low Low -150 -50 -100 -100 H MN -100 High High -200 0 -150 -50 I MN -75 High Medium -150 0 -100 -50 J MN -50 High Low -100 0 -50 -50 K MN -50 Low High -150 50 -100 0 L MN -25 Low Medium -100 50 -50 0 M MN 0 Low Low -50 50 0 0 M MP 0 Low Low -50 50 0 0 P MP 25 Low Medium -50 100 0 50 Q MP 50 Low High -50 150 0 100 R MP 50 High Low 0 100 50 50 S MP 75 High Medium 0 150 50 100 T MP 100 High High 0 200 50 150 U PP 100 Low Low 50 150 100 100 V PP 125 Low Medium 50 200 100 150 W PP 150 Low High 50 250 100 200 X PP 150 High Low 100 200 150 150 Y PP 175 High Medium 100 250 150 200 Z PP 200 High High 100 300 150 250 In each experiment, Trial M was presented to participants only once. However, for all analyses, the data for Trial M, which has an expected value of Zero, were included twice, once to represent the Mixed Negative set and once to represent the Mixed Positive set.
36 all positive (PP, gains). These levels are describe d in more detail in the Materials section. The dependent variable was Risk Preferen ce, which was measured as number of times out of 6 that the riskier option was selected. These 6 oppor tunities correspond to the six lottery pair trials within each Outcome Valence condition. To minimize potential order effects, eight different stimulus orders were used. These orders were systematically manipulated to ensure that each lottery pair would appear at the beginning, middle or end of the stimulus sequence for at least some participants. Also, the orders were checked to be sure that similar stimuli did not cooccur too often, and that the four outcome valence conditions were represented roughly evenly throughout the sequence of trials. Procedure Stimuli presentation and data collecti on were done via a computer program. Twelve computers in the laborat ory were set up so that two of those computers would run the same order of the lottery at any given time. Participants choosing to sit at those matched computers were therefore assigned to work as partners in a dyad. Previously unacquainted participants were assigned to dya ds in this way to re duce the likelihood that relationship dynamics could influence respons es. Matched computers were arranged such that the partners in a dyad would not face each other. Thus, care was taken to avoid the influence of nonverbal communicat ion. A minimum of four part icipants (two dyads) were run in each session to mainta in anonymity of partners. On entering the lab, participants were instru cted to take a seat at any of several computers with a lit screen. Instructions regarding the ba sic task followed, including
37 three practice lotteries which participants completed as individuals. Then participants were informed that they would complete the next segment of the study with an experimenter-assigned partner as part of a dyad. Each participant would be able to communicate with their assigned partner by Â“chattingÂ” (i.e., sharing of text messages) via an Instant Messenger (IM) window on their screen. Figure 4. Sample Screen Shot of a Trial in the Risky Choice Task. Dyads communicated through an Instant Messenger wi ndow (lower left), which remained on screen throughout the experiment. Here, Lo ttery 1 has two tickets -$200 and -$50, each of which has a 50% chance of being randomly drawn, resulting in a monetary loss. Similarly Lottery 2 has two tickets, -$150 a nd -$100, each of which has a 50% chance of being the outcome in a random draw of this loss lottery.
38 Dyads were instructed to use generic names (e.g., Â“ExpComp6Â”) pre-assigned by the experimenter to maintain anonymity. Dyads were encouraged to discuss their thoughts and reasons behind their impressions of one or the other lottery, and to Â“explain in details why you think a particul ar lottery is better than the other.Â” Dyads were also told that they must reach an agreement prior to choosing a lottery on the computer. In reality though each participant could select whichever lo ttery they preferred. (In practice, there was only one occasion in which a participant fail ed to select the lottery that the dyad had agreed to choose.) Both the Risky Choice and the Risk Mana gement task involved 23 decision trials. Figures 4 and 5 show sample screenshots for lo ttery presentations in the risky choice task and risk management tasks respectively. Risky choice task. On each trial in the risky choi ce task, a pair of lotteries was presented on the screen. Each dyad then shar ed their impressions back and forth via IM about which lottery to choose until an agreement was reached. Then each individual responded by clicking on the button beneath th e preferred lottery c onsistent with that agreement. Then the next lottery pa ir appeared, signali ng the next trial. Risk management task. In the risk management task, only one lottery was presented on the screen in each trial. Partic ipantsÂ’ task was to improve the outcome value of one of the two lottery ticke ts by $50. Dyad partners shar ed their impressions with one another via IM regarding which ticket to improve until an agreement was reached. Then each individual clicked on the agreed-upon tick et, which increased the face value of that potential outcome by $50 (and moved the ticket one column to the right on the lottery display). If the dyad selected the lower ticket to improve, the resulting lottery
39 corresponded to the low risk lottery (i.e., risk averse choice) from the comparable trial in the risky choice task. If the dyad selected the higher ticket to improve, the result corresponded to the high risk lott ery (i.e., risk seeking choice). Figure 5. Sample Screen Shot of a Trial in the Risk Management Task. Dyads communicated through an Instant Messenger wi ndow (lower left), which remained on screen throughout the experiment. The lott ery has two tickets -$50 and $50, Each ticket has a 50% chance of being randomly drawn. In this task, one of the two tickets can be moved to the next higher va lue on the number line. Moving the low ticket results in a less risky lottery with outcomes of $0 and $50, wh ereas moving the high ticket results in a riskier lottery with outcomes -$50 and $100. After completing the 23 trials of the decision making task, each participant independently answered a set of five openended questions presented on their computer.
40 These questions were designed to explore thei r thoughts and feelings regarding the nature of the task and experience of working with another person.
41 Results Data from the four studies were combined and analyzed using Repeated Measures Analysis of Variance feature of SPSS (PAS W Â– Version 18) in a 2x2x4 (Decision Maker [individual, dyad] x Task [risky choice, ri sk management] x Outcome Valence [all negative, mixed negative, mixed positive, a ll positive]) factorial design. Decision Maker was a between-subjects variab le representing whether the de cisions were being made by an individual or a dyad As mentioned before, the data for dyads were collected as part of the MasterÂ’s thesis, and compared to existing data for individuals (which were collected previously by Schneider, Hudspeth, & Decker, manuscript in preparation). Task was also a between-subjects variable; some of the participants completed the Risky Choice task and the remainder completed the Risk Management task. Outcome Valence was a withinsubjects variable with 4 levels: all negative (NN, losses ), mixed negative (MN), mixed positive (MP) and all positive (PP, gains ). The dependent variable measure of Risk Preference was defined as the number of times out of six that the riskier option was selected over the safer one. Analysis revealed a significant main eff ect of outcome valence (See Figure 6) on risk preference, F (3, 540)=55.07, p<0.05. The pattern of preferences does not clearly follow either PT or SP/A theory, though it has aspects of both. Consistent with prospect theory, there seems to be slightly more risk taking on average for loss lotteries than gain lotteries, especially in the mi xed conditions. However, consistent with SP/A theory, risk
42 preferences were generally on th e cautious side; a risk seek ing majority was not observed in any condition. Neither the main effect of Task or D ecision Maker was significant. Decision preferences for both the risky choice ( M =2.53, SD =1.91) and the risk management ( M =2.27, SD =1.96) tasks were neutral or slightly on the cautious side, F (1, 180)=1.32, n.s. Similarly, overall individuals ( M =2.49, SD =1.91) and dyads ( M =2.17, SD =1.98) Figure 6. Effect of Outcome Va lence on Risk Preference engaged in similar levels of risk taking, F (1, 180)= 2.69, n.s. Although the main effects were not significant, the impact of these variables was evident in interactions. The Outcome Valence x Task interaction was significant, F (3, 540)=49.7, p<0.05. (See Figure 7). As expected, there was a cros s-over interaction. In the risky choice task, preferences were risk averse fo r gains, whereas preferences we re slightly risk seeking for losses. The decision preference pattern in risky choice was consistent with Prospect 0 1 2 3 4 5 6 All NegativeMixed Negative Mixed Positive All PositiveRisk PreferenceOutcome ValenceEffect of Outcome Valence on Risk Preference Outcome Valence Risk Seeking Risk Aversion
43 TheoryÂ’s Value Function. Notably though, degree of risk seeking in the negative domain was at best modest. For risk management, the picture was consistent with SP/A theory predictions. Preferences tended to be risk averse for all outcome conditions except for Figure 7. Effect of Outcome Valenc e and Task on Risk Preference lotteries that guarantee some gain. There, pa rticipants were more willing to take a risk, though still not clearly on the side of majority risk taking. Follow up tests demonstrated that differences in risk preferences for the tw o tasks were statistically significant at every level of outcome valence. The Outcome Valence x Decision Maker interaction was also significant, F (3, 540)=3.33, p<0.05. As can be seen in Figure 8, th e overall pattern of ri sk preferences was similar at most outcome valences for indivi duals and dyads. The one exception was for the all negative lotteries, which imply loss of some amount for sure. When faced with this situation, dyads tended to take fewer ri sks than individuals, a nd pairwise comparison revealed that this difference was statistically reliable, F (1, 182)=8.31, p <0.05. For the 0 1 2 3 4 5 6 All NegativeMixed Negative Mixed Positive All PositiveRisk PreferenceOutcome ValenceOutcome Valence X Task Interaction Risky Choice Risk Management Risk Seeking Risk Aversion
44 other three outcome valence conditions, risk preferences between dyads and individuals were not significantly different. Thus, overa ll, it appears that dyadsÂ’ are more cautious than individuals in conditions in which losses are guaranteed, but there was little evidence of systematic differences elsewhere. Figure 8. Effect of Outcome Valence a nd Decision Maker on Risk Preference There was no Task x Decision Maker interaction, F (1,180)=0.13, n.s .; nor was there a three-way interaction of va lence, decision maker, and task, F (3, 540)=0.94, n.s. As can be seen in Figure 9, patterns conf irm the two-way interactions observed for outcome valence x task and outcome valence x decision maker. Risk preference patterns appear roughly consistent with Prospect Theory predictions for risky choice (Figure 9a) and roughly consistent with SP/A Theory pred ictions for risk management (Figure 9b). Dyads and individuals made similar decisions except for guaranteed losses in both tasks, when dyads were less willi ng to take risks. 0 1 2 3 4 5 6 All NegativeMixed Negative Mixed Positive All PositiveRisk PreferenceOutcome Valence Outcome Valence X Decisi on Maker Interaction Individual Dyad Risk Seeking Risk Aversion
45 Figure 9. Effect of Outcome Valence, Task and Decision Maker on Risk Preference Thus, the overall pattern of results su ggests some support for Prospect Theory, particularly for the risky c hoice task, and some support for the SP/A Theory, particularly for the risk management task. There was also some evidence that making decisions with a partner influences willingness to take risk, but only when losses were involved. 0 1 2 3 4 5 6All Negative Mixed Negative Mixed Positive All PositiveRisk PreferenceOutcome ValenceFig. 9a. Risky Choice Task Individual Dyad Risk Seeking Risk Aversion 0 1 2 3 4 5 6All NegativeMixed Negative Mixed Positive All PositiveRisk PreferenceOutcome ValenceFig. 9b. Risk Management Task Individual Dyad Risk Seeking RiskAversion
46 Discussion The present work explored the processes of risky decision making in individuals and dyads to see whether dyad decision might be subject to social or situational influences. Evidence was examined for the possi bility of risky versus cautious shifts, or for group polarization effects. Two different tasks were used in order to see whether changes in the decision context would produ ce different risk-taking strategies. The standard passive risky choice task was exp ected to engage shallower, psychophysical processes, whereas the active risk management task was expected to enlist more selfrelevant motivational and dis positional processes. As predicted, the risky choice task led to preferences consistent with prospect theory, and the risk management task led to preferences consistent with SP/A theory. Thus neither perspective coul d predict overall preference patte rns, but each of the two theoretical perspectives was su ccessful in a different task se tting. There was also some evidence of the role of social influences in that dyads tended to be more conservative than individuals in their deci sion behavior when dealing with undesirable outcomes. So a cautious shift was observed, but only for lo tteries involving losses. Theoretical implications of these findings are discussed next. Psychophysical Influences on Risky Decision Making Prospect Theory Value Function (Kahnema n and Tversky, 1979) predicts a risk preference pattern of risk aversion for gains and risk seeking for losses. Additionally losses are likely to be experienced as stronge r compared to gains of comparable values.
47 Also for both outcome valence domains, marg inally decreasing sensitivity would be expected. Overall, in the present work, Prospect Theory predicted results for Risky Choice for both individuals and dyads. Howe ver, Prospect TheoryÂ’s psychophysical explanations could not successfully account fo r risk decision prefer ences exhibited by individuals and dyads in the risk management task. It appears that for the passi ve risky choice task, the psychophysical and perceptual aspects of Prospect Theory are sufficient to predict behavior patterns. Change of reference from gains to losses appeared to infl uence participants in the Prospect Theory predicted direction such that they were genera lly risk averse for gains and risk seeking for losses, though not as risk s eeking as might be expected. Further evidence of psychophysical influen ces can be found in subtleties of the risky choice preference pattern across outcome valences. Decreasing marginal sensitivity predicts that reactions should get smaller as outcome va lue differences move away from zero. Careful review of Figure 7 shows that participants exhibited higher degrees of risk aversion and risk seeking respectively when responding to mixed lotteries for gains and losses compared to sure gains and sure losses Because the outcomes in the all negative and all positive conditions were further away from zero, this decrease in sensitivity to outcome differences may have w eakened risk propensities. Thus, it seems that a relatively peripheral level of processing may be adequate to make passive choices between given alternatives. However, preference patterns for the risk management task were markedly diffe rent, and could not be understood based on Prospect TheoryÂ’s psychophysical predictions. This could be because the nature of the
48 risk management task required a greater enga gement of mental pr ocesses bringing into focus motivational and dispos itional considerations. Dispositional and Motivational Influe nces on Risky Decision Making The Security-Potential/Aspiration Level Th eory (Lopes, 1987) predicts an overall risk averse behavior pattern for most indi viduals. Dispositionally, most people are expected to be more inclined towards secur ity (valuing safety more) though some may be more inclined toward potential (valuing opportu nities more; Schneider and Lopes, 1986). According to Lopes, when that sense of secu rity is threatened, mo st people would avoid risk. SP/A Theory also considers the impact of a situational factor, aspiration level, on risky decision behavior. On occasions in wh ich no threat to security is experienced, aspiration levels might be raised, and the in dividual could potentia lly aim for something better than the sure thing, or the comparable outcomes, choosing in favor of the riskier alternative. In situations of loss, risks will typically be avoided, with a few exceptions involving cases in which speci fic aspirations encourage c onsideration of the riskier alternative. In situations in which losses c ould potentially be avoided (mixed negative), occasional risk taking may occur in an effort to maintain the possibility of breaking even. Also, in cases of desperation, people may take risks as the only means of potential survival. In these cases, Weber and Millim an (1997) have found evidence that people reinterpret the technically ri skier option as safer one in these kinds of cases. This pattern of generally risk-ave rse preferences wa s observed for both individuals and dyads in the risk management task. For positive (gain) lotteries, where no possibility for loss existed, security was not threatened, and pe ople could potentially
49 have placed more weight on raising their aspiration levels, and meeting those levels. Therefore, compared to other outcome valence conditions, some risk taking was noticeable for the all positive lotteries. We hypothesized that this task, relative to risky choice, would be more likely to engage subj ects, and thus, to involve motivational and dispositional processes in a ddition to psychophysical ones. Indirect evidence for higher engagement in the risk management task comes from a preliminary coding analysis of the Inst ant Messenger conversations between dyad partners. About one-third of the participants in the risk management task expressed some concern with financial responsibility (a bout 60% of which focused on losses). Statements typically concerned Â“saving mone yÂ” or Â“lowering owed money.Â” The issue of financial responsibility was never brought up in the conversations between dyad partners in the risky choice task. This is probably another indication that the risk management task, though identical to the risky choice task in terms of stimulus materials, may have brought into focus motivational fact ors having to do with action and personal accountability. The opportunity for manipula ting outcome values may potentially have made participants more aware of their role s in the decision making process, enhancing motivation and a sense of responsibil ity for possible consequences. Social Influences on Risky Decision Making The risky and cautious shift perspectives pr edict a shift in groupÂ’s (in the present study, dyadÂ’s) responses in the risky or cautious direction respectively when compared to individualÂ’s decisi on preferences. Group polarizati on concepts on the other hand, suggest that dyads are likely to exhibit similar patterns of behavior as majority of individuals; only, for groups, th ese patterns are likely to be more extreme. This means
50 that if majority of individuals were risk seekers to begin with, dyads would engage in even more risk taking, whereas if individuals were risk averse initially, dyads would exhibit more pronounced risk aversion tendencies. Results do not show reliable differences between individual and dyad preferences in all cases. Only for sure losses, dyads seem to have engaged in significantly less risk taking compared to individuals. This raises th e possibility that when the direct threat to security was greatest, it became powerful enough in a social context to discourage risk taking. In terms of cautious shift versus gr oup polarization, we can examine the results to see which is more likely to have occurred. If security motivation is what is important, then a cautious shift explan ation would be supported. Given that the preference patterns for risky choice and risk management were opposite for the all negative condition, where individual-dyad differences were found, we can readily assess which explanation is most fitting. Individuals exhibited a risk seeking majority in risky choice, but a risk aver se majority in risk management. Group polarization would predict that the dyads woul d engage in even grea ter risk seeking in risky choice and greater risk aversion for risk management. A cautious shift explanation would predict that dyads would be more risk averse than individuals in both tasks. The cautious shift explanation was supported because as Figure 9 shows, the dyads were less willing to take risks in the allnegative condition for both tasks. This is consistent with the possibility that security motivation tended to be stronger in inte racting dyads than it was for individuals. Nevertheless, this ca nnot explain why dyads were not more risk averse overall.
51 Marquis and Reitz back in 1969 obtained si milar findings in a study the effects of uncertainty and group discussion on individualsÂ’ willingness to take risks. Marquis and Reitz (1969) found cautious shifts, but only for negative expected values (comparable to losses. However, they also found a risky sh ift for positive expected values (comparable to gains). Both of these findings are consiste nt with an SP/A explan ation. Perhaps with more power, this type of pattern would also have been observed in the current study. The conversational data are also some what supportive of a security-based description of dyad-individual differences. A preliminary coding analysis of the Instant Messenger conversations showed that dyad participants, in general, tended to talk more about losses than anything else. They also exhibited much concern with worse outcomes, and this was particularly evident in the risk management task. All negative lotteries presented decision situations where participan ts experienced a combined impact of both undesirable outcomes. Therefore, it is possibl e that dyads felt most vulnerable in these situations involving all negative lotteries, a nd wanted to reduce the unpleasant experience of threat by avoiding risk, as best as they could, given the circumstances. One interesting point made by some part icipants came at the end of the study, when all participants were asked to re spond to several questi ons regarding their experiences about working in dyads. In response to a question about whether the individual participant would have made d ecisions differently had they been working alone (as compared to working as part of a dyad), some of them brought up the point of experiencing some concern for the partnerÂ’s welfare. One sample comment was: Â“Â… since I had a partner, I tried to think of him/her so he/she wouldn't lose money and act[ed] more conservativelyÂ”. Thus, when working in a dyad context, participants
52 expressed discomfort with risking the loss of other peopleÂ’s money. This could be one reason why many participants when working as a dyad tended to err on the side of caution rather than risk in the negative domain. Limitations and Future Directions This thesis provides an initial exploration into situational and social influences on decision making. In the future, these variables might be combined into one larger formal experiment with random assignment of particip ants in order to re plicate these findings with greater control of extran eous variables. In addition, recruiting larger samples of participants would afford additional statis tical power to reliably document effects of moderate size, especially given the necessity to exclude participants who do not show a basic understanding of probabi lities or lotteries. Future studies might benefit from including a priori measures of risk attitude or dispositions. However, the current findi ngs suggest that th is may not be as straightforward as some might imagine. We found that risky d ecision making patterns differ with situations, suggesting that an in dividual may not have a single overarching risk attitude or disposition that transcends situations. For example, an individual may have conservative views towards financial i nvestments and yet be an adventure sports enthusiast, indicating that risk dispositions ma y vary with respect to different domains or situations in real life. In a ddition, even the interpretation of riskiness may not always be a point of agreement across individuals, or between researchers and participants. Future research comparisons might also go beyond individual versus dyad to include other comparisons that could bring in teams and/or groups of different types and
53 sizes. Exploring different types of social interactions as well as different settings for those interactions may be key to understa nding real world decision making. Finally, the present work suggests the need for an overarching theory to connect the gap between psychophysical, motivational, dispositional and social (not to mention cognitive) processes to successfully predict general patterns of ri sky decision behavior across decision situations and social settings. One promising direction has been taken in dual process perspectives of decision making. It has b een increasingly accepted by decision theorists and research ers that risky decision maki ng processes may involve two different systems of processing of inform ation (Damasio, 1994; Epstein, 1994; Evans, 2010; Finucane, Peters, & Slovic, 2003; Hogarth, 2001; Kahneman, 2003; Lieberman, 2000; Reyna, 2004; Stanovich & West, 2000). One system is hypothesized to involve relatively superficial pr ocesses, which are heuristic-base d and involve easy or automatic rules for quickly responding to familiar situ ations. The other system involves deeper conscious processes, which are an alytical in nature and involve a systematic evaluation of the decision situation. The nature of task si tuations may play a role in determining how these systems interact, and therefore influe nce both the decision process and resulting behavior. The risky choice task, for example, may be one situation in which heuristic approaches may be sufficient to make risk-related decisions, whereas the risk management task may be a situation in which a different heuristic or a more analytical approach is needed. There may be many different default heuristics depending on the type of decision situation, assuming that the situation feels familiar. For unfamiliar situations, a more analytic approach ma y be needed (e.g., Stanovich & West, 2000).
54 Social situations may place additional constrains on decision strategies. In real life, there are many different kinds of decision situations th at are likely to be more or less familiar. Determining how the tradeoff between these two systems occurs may help us resolve when these different theories may come into play. Advances in neuroscience are likely to be particularly informative in making progress along these lines. Decision researchers may sometimes seem to forget that in real life, we often make decisions with different people in diffe rent sets of circumst ances and that these characteristics define decisi on situations. A comprehens ive understand ing of human decision making processes requires sensitivity to the many different kinds of social and contextual influences that may be at wo rk in different decision situations.
55 References Allport, H. (1924). Social Psychology. Boston, MA: Houghton Mifflin. Aron, A., Norman, C., Aron, E., McKenna, C ., & Heyman, R. (2000). Couples' shared participation in novel and arousing activiti es and experienced relationship quality. Journal of Personality and Social Psychology 78 (2), 273-284. Bernoulli, D. (1954). Exposition of a new theory on the measurement of risk. Econometrica: Journal of the Econometric Society 22 (1), 23Â–36. Bostrom, A., Hoffmann, S., Krupnick, A., Adamowicz, W., Goldman, R., McWilliams, M., & Varner, J. (2005, November). Pare ntal decision-making and children's health. Working paper presented at EPA NCER Star grantees workshop, Washington, DC. Brown, R. (1965). Social Psychology. New York: The Free Press. Burnstein, E., & Vinokur, A. (1977). Persuasive argumentation and social comparison as determinants of atti tude polarization. Journal of Experime ntal Social Psychology 13 (4), 315-332. Camerer, C. (2003). Behavioral game theory: Experime nts in strategic interaction New York, NY US: Russell Sage Foundation. Cartwright, D. (1971). Risk ta king by individuals and groups: An assessment of research employing choice dilemmas. Journal of Personality and Social Psychology 20 (3), 361-378.
56 Charles, C., Gafni, A., & Whelan, T. ( 1997). Shared decision-making in the medical encounter: What does it mean? (Or it takes at least two to tango). Social Science & Medicine 44 (5), 681-692. Clark, M. (1984). Record keeping in two types of relationships. Journal of Personality and Social Psychology 47 (3), 549-557. Crowe, E., & Higgins, E. T. (1997). Regul atory focus and strategic inclinations: Promotion and prevention in decision-making. Organizational behavior and human decision processes 69 117Â–132. Damasio, A. R. (1994). DescartesÂ’ error: Emotion, reason, and the human brain New York: Putnam. Deets, M., & Hoyt, G. (1970). Variance Pr eferences and Variance Shifts in Group Investment Decisions. Organizational Behavior & Human Performance 5 (4), 378-386. Elster, J. (1989). Social Norms and Economic Theory. Journal of Economic Perspectives, 3 (4) 99Â–117. Epstein, S. (1994). Integration of th e cognitive and psychodynamic unconscious. American Psychologist, 49, 709-724. Evans, J. S. B. T. (2008). Dual-processing accounts of reasoning, judgment, and social cognition. Annual Review of Psychology, 59, 255Â–278. Farnsworth, P. R., & Behner, A. (1931). A not e on the attitude of social conformity. Journal of Social Psychology, 2 126-128. Fehr, E., & Fischbacher, U. (2004). So cial norms and human cooperation. Trends in Cognitive Sciences 8 (4), 185-190.
57 Finucane, M. L., Peters, E., & Slovic, P. (2003). Judgment and decision making: The dance of affect and reason. In S. L. Schneider & J. Shanteau (Eds.) Emerging perspectives on judgment and deci sion research, Part III (pp. 327-364). New York: Cambridge University Press. Fiske, A. P. (1992). The Four Elementary Forms of Sociality: Framework for a Unified Theory of Social Relations. Psychological Review, 99 (4), 689-723. Frosch, D., & Kaplan, R. (1999). Shared deci sion making in clinical medicine: Past research and future directions. American Journal of Preventive Medicine 17 (4), 285-294. Gardner, M., & Steinberg, L. (2005). Peer Influence on Risk Taking, Risk Preference, and Risky Decision Making in Adolescen ce and Adulthood: An Experimental Study. Developmental Psychology 41 (4), 625-635. Hardisty, D. J. & Sanitioso, R. B. (2007, Ma y). Text Based Discussion Influences Group Risk Preference and Decision Making. A ssociation for Psychological Science, Washington, DC. Hershey, J., & Schoemaker, P. (1980). Pros pect Theory's Reflection Hypothesis: A Critical Examination. Organizational Behavior & Human Performance 25 (3), 395-418. Heyman, J., & Ariely, D. (2004). Effort for Payment. Psychological Science (WileyBlackwell) 15 (11), 787-793. Higgins, E. T. (1987). Self-discrepancy: A theory relating self and affect. Psychological Review 94 (3), 319-340.
58 Higgins, E. T. (1996). The Â“self digestÂ”: Self-knowledge serv ing self regulatory functions. Journal of Personality and Social Psychology 71 1062Â–1083. Higgins, E. T. (1997). Beyond pleasure and pain. American Psychologist 52 (12), 12801300. Higgins, E. T. (1998). From expectancies to worldviews: Regulatory focus in socialization and cognition. Attribution and social interaction: The legacy of Edward E. Jones (pp. 243-309). Washington, DC US: American Psychological Association. Higgins, E. T., Roney, C. J. R., Crowe, E., & Hymes, C. (1994). Ideal Versus Ought Predilections for Approach and Avoidan ce: Distinct Self-Regulatory Systems. Journal of Personality and Social Psychology, 66 (2), 276-286. Hogarth, R. M. (2001). Educating intuition Chicago: University of Chicago. Hunt, E. B., & Rowe, R. R. (1960). Group a nd individual economic decision making in risk conditions. In D. W. Taylor (Ed.), Experiments on decision making and other studies (pp. 21-25). Arlington, VA.: Armed Se rvices Technical Information Agency. Isenberg, D. J. (1986). Group polarization: A critical review and meta-analysis. Journal of Personality and Social Psychology 50 (6), 1141Â–1151. Kahneman, D. (2003). A perspective on judgment and choice: Mapping bounded rationality. American Psychologist, 58 697-720. Kahneman, D., & Tversky, A. (1979). Prospect Theory: An Analysis of Decision Under Risk. Econometrica, 47 (2), 263-91.
59 Keller, L., Sarin, R., & Sounderpandian, J. (2007). An Examination of Ambiguity Aversion: Are Two Heads Better Than One? Judgment and Decision Making 2 (6), 390-397. Kelley, H. H., & Thibaut, J. W. (1954). Expe rimental studies of group problem solving and process. In G. Lindzey (Ed.) Handbook of Social Psychology Cambridge: Addison-Wesley. Kerr, N., Garst, J., Lewandowski, D., & Ha rris, S. (1997). That still, small voice: Commitment to cooperate as an internalized versus a social norm. Personality and Social Psychology Bulletin 23 (12), 1300-1311. Kocher, M. G., & Sutter, M. (2005). The D ecision Maker Matters: Individual Versus Group Behaviour in Experimental Beauty-Contest Games. The Economic Journal 115 (500), 200Â–223. Khberger, A. (1998). The Influence of fram ing on risky decisions: A meta-analysis. Organizational Behavior and Hu man Decision Processes,75 23Â–55. Laurenceau, J., Barrett, L., & Pietromonaco, P. (1998). Intimacy as an interpersonal process: The importance of self-disclos ure, partner disclosure, and perceived partner responsiveness in interpersonal exchanges. Journal of Personality and Social Psychology 74 (5), 1238-1251. Lgar, F., Stacey, D., Graham, I., Elw yn, G., Pluye, P., Gagnon, M., et al. (2008). Advancing theories, models and measuremen t for an interprofessional approach to shared decision making in pr imary care: a study protocol. BMC Health Services Research 8, 1-8.
60 Levin, I. P., Schneider, S. L., & Gaeth, G. J. (1998). All frames are not created equal: A typology and critical analys is of framing effects. Organizational behavior and human decision processes 76 149Â–188. Lieberman, M.D. (2000). Intuition: A so cial cognitive neuroscience approach. Psychological Bulletin, 126 109-137. Lopes, L. (1984). Risk and distributional inequality. Journal of Experimental Psychology: Human Perception and Performance 10 (4), 465-485. Lopes, L. (1987). Between hope a nd fear: The psychology of risk. Advances in experimental social psychology, Vol. 20 (pp. 255-295). Luhan, W., Kocher, M., & Sutter, M. (2009). Group Polarization in the Team Dictator Game Reconsidered. Experimental Economics 12 (1), 26-41. Marquis, D., & Reitz, H. (1969). Effect of uncertainty on risk taking in individual and group decisions. Behavioral Science 14 (4), 281-288. McCauley, C., Stitt, C. L., Woods, K., & Lipt on, D. (1973). Group shift to caution at the race track. Journal of Experimental Social Psychology, 9, 80-86. Mellers, B., Schwartz, A., & Cooke, A. (1998). Judgment and decision making. Annual Review of Psychology 49, 447-477. Mellers, B., Schwartz, A., & Weber, E. (1997). Do risk attitudes refl ect in the eye of the beholder?. Choice, decision, and measurement: Essays in honor of R. Duncan Luce (pp. 57-71). Mahwah, NJ US: Lawren ce Erlbaum Associates Publishers. Moscovici, S., & Zavalloni, M. (1969). Th e group as a polarizer of attitudes. Journal of Personality and Social Psychology 12 (2), 125-135.
61 Myers, D., & Lamm, H. (1976). The group polarization phenomenon. Psychological Bulletin 83 (4), 602-627. Nordhy, F. (1962). Group interaction in decision making under risk (Unpublished MasterÂ’s thesis). Massachusetts Inst itute of Technology, School of Industrial Management, Cambridge, MA. Rabow, J., Fowler, F. J., Bradford, D. L., Ho feller, M. A., & Shibuya, Y. (1966). The role of social norms and leadership in risk taking. Sociometry, 29, 16-27. Reyna, V. F. (2004). How people make decisi ons that involve risk: A dual-processes approach. Current Directions in Psychological Sciences, 13 60-66. Savage, L. (1954). The foundations of statistics Oxford, England: John Wiley & Sons. Schneider, S., & Lopes, L. (1986). Reflecti on in preferences under risk: Who and when may suggest why. Journal of Experimental Ps ychology: Human Perception and Performance 12 (4), 535-548. Shah, J., & Higgins, E. T. (1997). Expectan cy x value effects: Regulatory focus as determinant of magnitude and direction. Journal of Personality and Social Psychology 73 (3), 447-458. Stanovich, K. E., & West, R. F. (2000). Individual differences in reasoning: Implications for the rationality debate? Behavioral and Brain Sciences, 23 645-726. Stasser, G., & Titus, W. (2006). Pooling of Unshared Information in Group Decision Making: Biased Information Sampling During Discussion. Small groups (pp. 227239). New York, NY US: Psychology Press.
62 Stoner, J. A. F. (1961). A Comparison of individual and gr oup decisions involving risk (Unpublished masterÂ’s thesis). Massac husetts Institute of Technology, Sloan School of Management, Cambridge, MA. Stoner, J. A. F. (1968). Risky and cautious shifts in group decisi ons: The influence of widely held values. Journal of Experimental Social Psychology, 4, 442-459. Towle, A., & Godolphin, W. (1999). Framewor k for teaching and learning informed shared decision making. British Medical Journal, 319, 766-771. Tversky, A. & Kahneman, D. (1981). The fram ing of decisions and the psychology of choice. Science, 211, 453Â–458. Tversky, A., & Kahneman, D. (1991). Loss Av ersion in Riskless Choice: A ReferenceDependent Model. The Quarterly Journal of Economics, 106 (4), 1039-1061. Vinokur, A., & Burnstein, E. (1974). The effects of partially shared persuasive arguments on group-induced shifts: A gr oup-problem-solving approach. Journal of Personality and Social Psychology, 29, 305-315. Vinokur, A., & Burnstein, E. (1978). Depol arization of atti tudes in groups. Journal of Personality and Social Psychology 36 (8), 872-885. Vinokur, A., & Burnstein, E. (1978). Novel argu mentation and attitude change: The case of polarization following group discussion. European Journal of Social Psychology 8 (3), 335-348. Vinokur, A., Trope, Y., & Burnstein, E. (1975). A decision-making analysis of persuasive argumentation a nd the choice-shift effect. Journal of Experimental Social Psychology 11 (2), 127-148.
63 Von Neumann, J., & Morgenstern, O. (1944). Theory of games and economic behavior Princeton, NJ US: Princeton University Press. Von Neumann, J., & Morgenstern, O. (1947). Theory of games and economic behavior (2nd Rev. ed.) Princeton, NJ US: Princeton University Press. Wallach, M., Kogan, N., & Bem, D. (1962).Gro up influence on indivi dual risk taking. The Journal of Abnormal and Social Psychology 65 (2), 75-86. Wallach, M., Kogan, N., & Bem, D. (1964). Diffu sion of responsibility and level of risk taking in groups. The Journal of Abnormal and Social Psychology 68 (3), 263274. Weber, E. U., & Milliman, R. A. (1997). Perceived Risk Attitudes: Relating Risk Perception to Risky Choice. Management Science, 43 (2), 123-144. Weber, E.U. (1999). WhoÂ’s afraid of a little risk? New evidence for general risk aversion. In J. Shanteau, B.A. Mellers, & D.A. Schum (Eds.), Decision science and technology: Reflections on the c ontributions of Ward Edwards (pp. 53-64). Norwell, MA: Kluwer Academic Publishers. Wood, W. (2000). Attitude change: Pe rsuasion and social influence. Annual Review of Psychology 51 539-570. Zaleska, M. (1974). The effects of disc ussion on group and individual choices among bets. European Journal of Social Psychology, 4, 229-250. Zaleska, M. (1976). Majority infl uence on group choices among bets. Journal of Personality and Social Psychology, 33, 8-17.