UNDERSTANDING THE ROLE OF RISK AND UNCERTAINTY IN SEQUENTIAL
DECISION MAKING DURING INFORMATION ACQUISITION
By
Michael Thomas Braun

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Psychology
2012

ABSTRCT
UNDERSTANDING THE ROLE OF RISK AND UNCERTAINTY IN SEQUENTIAL
DECISION MAKING DURING INFORMATION ACQUISITION
By
Michael Thomas Braun
When employees are faced with novel situations it is often necessary for them to acquire
new information. Unreliability of information sources and uncertainty regarding information
usefulness make the information acquisition process analogous to a sequence of risky decisions.
Prospect Theory (Kahneman & Tversky, 1979) and its generalizations are currently the dominant
approach to the study of risky decision making in psychology and economics due to their ability
to accurately describe the decision biases that are frequently document in the presence of risk
and uncertainty. Using a lab-based computer simulation, CRONUS SOLO, common biases from
the Prospect Theory literature are examined for sequential risky decision making during
information acquisition. Additionally, common relationships between individual differences and
sequential risky decision making as well as the effect of learning and feedback on risky decision
making are explored. The biases present during the information acquisition process are not fully
explained by Prospect Theory and its generalizations. Overall, individuals became more
efficient and optimal decision makers over time. The effects of information reliability,
information framing, and specific individual differences are presented. Finally, potential
explanations for the current findings as well as implications for training are discussed.

ACKNOWLEDGEMENTS

I would like to extend my extreme gratitude to my advisor, Dr. Steve Kozlowski, for all
of the time, effort, and support that he has given me throughout this research process. I am
undoubtedly a better researcher, writer, and presenter thanks to his guidance.
I would also like to thank my committee members, Rick DeShon, Neal Schmitt and
Georgia Chao, for providing excellent guidance and insight on improving my research.
Finally, thanks to all of my friends and family for all of the loving support they have
always provided me. A special thanks to my parents for their continued love, encouragement,
patience, and guidance.

Michael T. Braun gratefully acknowledges the Office of Naval Research (ONR),
Command Decision Making (CDM) Program (N00014-09-1-0519, S.W.J. Kozlowski and G. T.
Chao, Principal Investigators) for support of this research. Any opinions, findings, and
conclusions or recommendations expressed are those of the author and do not necessarily reflect
the views of ONR.

iii

TABLE OF CONTENTS

LIST OF TABLES ......................................................................................................................... vi
LIST OF FIGURES ..................................................................................................................... viii
INTRODUCTION ...........................................................................................................................1
Organizational Decision Making .........................................................................................5
Cognitive Decision Making .....................................................................................6
Behavioral Decision Making ...................................................................................8
Limitations of Organizational Decision Making ...................................................14
Prospect Theory: Describing Risky Decision Making.......................................................16
Prospect Theory .....................................................................................................17
Cumulative Prospect Theory..................................................................................25
The House Money Effect and Escalation of Commitment .....................................28
Risky Decision Making during Information Acquisition ..................................................30
Prospect Theory Hypotheses..................................................................................33
Individual Differences Hypotheses ........................................................................37
Learning and Decision Making Hypotheses ..........................................................40
METHOD ......................................................................................................................................42
Sample ...............................................................................................................................42
Procedure ...........................................................................................................................42
CRONUS SOLO ...............................................................................................................43
Scenario Design .....................................................................................................51
Measures ............................................................................................................................53
Demographics ........................................................................................................53
Cognitive ability.....................................................................................................53
Affect .....................................................................................................................53
Goal Orientation.....................................................................................................54
Risk Propensity ......................................................................................................54
Time Decision Making ..........................................................................................54
Deviations from Optimal Decisions.......................................................................55
Risky Decision Making..........................................................................................55
Performance ...........................................................................................................56
Data Analysis ....................................................................................................................58
RESULTS ......................................................................................................................................60
Prospect Theory Hypotheses .............................................................................................60
Individual Differences Hypotheses ...................................................................................73
Learning and Decision Making Hypotheses .....................................................................81
DISCUSSION ................................................................................................................................85
Limitations and Future Directions .....................................................................................92
iv

Conclusion .........................................................................................................................96
APPENDIX A: Individual Differences Scales ..............................................................................98
REFERENCES ............................................................................................................................102

v

LIST OF TABLES

Table 1. Obstacle Properties in CRONUS SOLO ........................................................................53
Table 2. Correlations among Focal Variables...............................................................................57
Table 3. ICC(1) Values of Time-varying Covariates....................................................................58
Table 4. Average Proportion of High Probability Information Posted by Trial ...........................61
Table 5. Proportion of Individuals who engaged in each Category of Posting Behavior
for High Probability Information ....................................................................................62
Table 6. Differences in Proportion of High Probability Information Posted between
Conditions .......................................................................................................................64
Table 7. Average Proportion of Low Probability Information Posted by Trial ............................65
Table 8. Proportion of Individuals who engaged in each Category of Posting Behavior
for Low Probability Information .....................................................................................66
Table 9. Differences in Proportion of Low Probability Information Posted between
Conditions .......................................................................................................................66
Table 10. Average Proportion of Moderate Probability Information Posted by Trial ..................67
Table 11. Proportion of Individuals who engaged in each Category of Posting Behavior
for Moderate Probability Information...........................................................................68
Table 12. Differences in Proportion of Moderate Probability Information Posted between
Conditions .....................................................................................................................69
Table 13. Effect of Risky Decision Making on Performance .......................................................71
Table 14. Effect of Negative Affectivity on Risky Decision Making ..........................................74
Table 15. Effect of Negative Affectivity on Performance ............................................................74
Table 16. Effect of Positive Affectivity on Positive Information Attended .................................75
Table 17. Effect of Positive Affectivity on Risky Decision Making ............................................76
Table 18. Effect of Performance Avoid Goal Orientation on Risky Decision Making ................77

vi

Table 19. Effect of Performance Approach Goal Orientation on Risky Decision Making ..........78
Table 20. Effect of Risk Taking Propensity on Risky Decision Making......................................79
Table 21. Effect of Cognitive Ability on Risky Decision Making ...............................................80
Table 22. Effect of Cognitive Ability on Performance .................................................................80
Table 23. Effect of Time on Performance ....................................................................................82
Table 24. Effect of Time on Deviations from Optimal Probability ..............................................83
Table 25. Effect of Time on Time to Make Decisions .................................................................84

vii

LIST OF FIGURES

Figure 1. Hypothetical value function for Prospect Theory .........................................................23
Figure 2. Hypothetical weighting function for Prospect Theory ..................................................24
Figure 3. Hypothetical weighting function for Cumulative Prospect Theory ..............................27
Figure 4. Specialist Map screen for CRONUS SOLO. For interpretation of the references
to color in this and all other figures, the reader is referred to the electronic version
of this dissertation .........................................................................................................46
Figure 5. Mission Info screen for CRONUS SOLO .....................................................................47
Figure 6. Specialist Info screen for CRONUS SOLO ..................................................................48
Figure 7. Mission Command screen for CRONUS SOLO ...........................................................49
Figure 8. Feedback screen for CRONUS SOLO ..........................................................................50

viii

INTRODUCTION

Organizational success often hinges on the ability of individuals to make correct
decisions in the presence of high degrees of risk and uncertainty. Frequently, these decisions
affect not only the organization but the community at large as well. For example, the United
States intelligence community “gathers tremendous amounts of information based on a complex
set of needs and requirements established by the policy makers it supports. …Much of these data
are of questionable reliability and …is reviewed and evaluated by intelligence collectors and
analysts who gauge its reliability and accuracy” (Permanent Select Committee on Intelligence:
House of Representatives, 1996; p. 13). To properly identify relevant threats to national security,
the intelligence collectors and analysts must make accurate decisions on what information is
most reliable and what information is most important. Failure to do so can result in catastrophic
disasters such as the terrorist attacks on September 11, 2001. Similarly, emergency medical
personnel are bombarded with information on the patient they are treating such as medical
history, reported symptoms, and physiological responses. To correctly diagnose and treat a
patient, the individual must wade through the plethora of information and only attend to the data
that are most likely relevant and most critical. Understanding the decision making process that
individuals go through when presented with large quantities of uncertain, risky information is
vital to the success of the organizations like those just described. Unfortunately, to date there is
little research that directly studies this phenomenon.
Not surprisingly, given the impact that decision making can have on organizational
success, there is an abundance of organizational research devoted to it. Unfortunately, like many
areas of concentration in the organizational sciences, the study of decision making is plagued by

1

an overreliance on cross-sectional and/or self report data. Additionally, many of the more recent
studies of organizational decision making are focused at the team level. While understanding
how teams make decisions is certainly important given the shift of organizations to utilize teams
for problem solving, it is also necessary to understand how the building blocks of teams (i.e.,
individuals) integrate information to make decisions. Finally, most theories of organizational
decision making are built on the assumption of a deterministic environment, ignoring the
uncertain or stochastic nature of the real world. Therefore, to help organizations understand and
improve sequential decision making, it is necessary to examine how individuals make a series of
decisions over time in an uncertain environment with multiple risky outcomes.
The purpose of the current research is to understand the information acquisition process
by modeling it as a series of risky decisions. More specifically, this dissertation categorizes the
organizational decision making literature and highlights its shortcomings for understanding risky
decision making. Additionally, the predictions of Prospect Theory, the dominant approach to
risky decision making in other fields, are integrated with the type of decisions individuals make
during information acquisition. Finally, relationships among individual difference variables,
learning, and risky decision making that are commonly found in economics, cognitive
psychology, and the organizational sciences are tested for the decisions made during the
information acquisition process. Special attention is paid throughout the paper to the role of time
in risky decision making and information acquisition.
The study of decision making under the conditions of risk and uncertainty is common in
many scientific disciplines. In the social sciences, it is most frequently researched in economics
and cognitive psychology. Initially, individuals were thought to be rational decision makers and
were modeled accordingly with theories such as Expected Utility Theory (Atkinson, 1957;

2

Keeney & Raiffa, 1976). However, it was frequently documented that individuals deviate from
rational choice when presented with risky uncertain outcomes (e.g., Kahneman & Tversky, 1979;
Markowitz, 1952). As a result, an alternative class of models emerged which incorporated the
heuristics and biases that people use in decision making, thus, making them better descriptors of
actual behavior. Prospect Theory (PT) models decisions as a function of the relative gains and
losses of each outcome weighted by the perceived impact of the outcomes (Kahneman &
Tversky, 1979; Tversky & Kahneman, 1992). By incorporating perceptions and the unequal
weighting of relative gains and losses, PT better describes human behavior than the original
rational choice models. As a result, PT is currently the dominant approach to the study of
decision making under risk and uncertainty.
Despite the ability of PT to describe actual decision making, its focus and application has
been rather narrow in scope. In their initial presentation, Kahneman and Tversky state that
“[Prospect] Theory is developed for simple prospects with monetary outcomes and stated
probabilities” (1979; p. 274). Not surprisingly, then, the study of prospect theory has
predominantly been in financial disciplines such as finance and economics. Even when other
fields such as psychology have utilized the theory, it has primarily been done using monetary
decisions (e.g., gambles). Similarly, Prospect Theory was created to explain single decisions.
Little research exists that examines how the effect of learning or experience over a sequence of
decisions would impact the accuracy of PT to describe actual behavior.
Despite the initially intended purpose of PT, Kahneman and Tversky (1979) recognize
that “it can be extended to more involved choices” (p. 274). Recently, there have been
arguments that time is a valuable resource similar to money, and that spending time on
something can result in successes or failures equivalent to those in financial decisions (Cunha &

3

Caldieraro, 2009; Steel & Konig, 2006). Additionally, research has been conducted on possible
boundary conditions of the predictions of prospect theory for sequential decisions (e.g.,
escalation of commitment; Staw, 1976; house money effect; Thaler & Johnson, 1990). As a
result, PT and its generalizations provide a good framework to predict how individuals make a
sequence of decisions in an uncertain, risky environment.
When acquiring new information, employees must quickly decide what information to
attend to and what to ignore. Understanding this process is important because in organizations
like those described earlier, workers are frequently bombarded with an extreme volume of
information and the validity and usefulness of the information is not always certain. Choosing
whether to attend to or ignore each piece of information has potential costs and benefits
associated with it, and those costs and benefits each have a likelihood of occurrence. Since
workers encounter both risk and uncertainty when deciding what to attend to, the information
acquisition process becomes analogous to a sequence of risky decisions. While the
organizational decision making literature lacks strong theories of decision making under risk,
Prospect Theory and its generalizations provide a solid foundation for describing the risky
decision making process that employees engage in when acquiring new information.
To determine whether prospect theory can accurately predict information searching under
risky uncertain conditions, the paper is organized in the following manner: First, I review and
categorize the organizational decision making literature, highlighting its shortcomings for
understanding sequential decision making under risk and uncertainty. I then briefly define
prospect theory, its origin, primary components, and its mathematical representation, including
the generalization to cumulative prospect theory and the special cases of the house money effect
and escalation of commitment. Next, I utilize the predictions of Prospect Theory and

4

Cumulative Prospect Theory to formulate specific hypotheses about the biases that individuals
may demonstrate when engaging in the information search process. Additionally, common
findings from the Prospect Theory literature are used to formulate predictions about what
individual differences can affect the common biases in risky decision making. I also draw upon
findings from the house money effect and escalation of commitment to understand how prior
successes and failures could impact subsequent decision making. I then describe my sample,
procedure, computerized task, operationalizations of risk, uncertainty, and information
acquisition, and proposed data analytic strategy. Finally, I present results for each hypothesis
followed by a discussion section summarizing the results, highlighting their implications, and
discussing some limitations and future research directions.

Organizational Decision Making
Countless numbers of decisions are made in organizations each day. Not only do the
quality and efficiency of decisions affect vital organizational outcomes (Shrode & Brown, 1970;
Ziller, 1957) but they have also been shown to affect employees’ affective reactions to the
organization (Jackson, 1983, 1984; Ritche & Miles, 1970). There are numerous types of
decisions that individuals can engage in, leading the study of decision making to encompass a
wide range of scholarly work. The applicability of theories and prior research findings depends
on the type of decision under investigation. The organizational decision making literature can be
grouped into two broad categories, cognitive decision making and behavioral decision making.
Cognitive decision making encompasses studies that investigate decisions where individuals or
teams must complete some task of cognitive ability. These studies frequently operationalize
decision making as either creativity or performance. Conversely, behavioral decision making

5

defines studies where individuals must decide on an appropriate behavior to solve a particular
problem, frequently without an obvious “correct” solution. Behavioral decision making
incorporates a wide variety of decision types, including strategic decision making, dynamic
decision making, open-ended choice problems, discrete choice problems, and gambles with
monetary and nonmonetary outcomes. Each type of cognitive and behavioral decision making
has unique properties related to the type of decision under investigation. As such, each type will
be described and relevant relationships with organizational variables will be explained, followed
by an overview of the limitations of the current organizational decision making literature to
describe sequential decision making under risk and uncertainty.

Cognitive Decision Making
Cognitive decision making in the organizational sciences is frequently operationalized in
two distinct ways. The first is by creativity tasks where individuals or teams are asked to make a
list of all possible solutions to some question. The greater number of solutions the individual or
team can generate, the higher the performance on decision making. Frequently, the quality of the
potential solutions generated is completely ignored. A number of important constructs have been
found to impact the quality of the creative decision making process. At the individual level, it
was found that stress and time pressure inhibit the decision making process and fewer ideas are
generated. Additionally, when questions are framed in an open-ended and ambiguous manner,
individuals had greater difficulty effectively framing the problem and generating plausible
solutions (Abualsamh, Carlin, & McDaniel, 1990). It has also been documented that initially
suppressing emotions, so as to not interfere with the decision making process, had a positive
effect on creativity (Guzzo & Waters, 1982).

6

At the team level, communication was found to be instrumental in increasing creative
decision making in a number of important ways. The medium by which the team communicates
is important, with face to face communication being more effective than computer mediated for
idea generation. Along the same lines, as tasks become more interdependent, high quality teammember exchange (TMX) increases the decision making ability of the team (Alge, Wiethoff, &
Klein, 2003). Additionally, during communication, the dissent of a minority opinion increases
creativity and, thus, decision making (De Dreu & West, 2001). Finally, imposing structure on
the communication of the group by forcing team members to first generate ideas independently
and then discuss them as a group led to significantly better decision making processes (van de
Ven & Delbecq, 1974).
The second type of cognitive decision making paradigm in the organizational sciences is
to study decision making as performance. Given that performance is highly contextualized, there
are as many ways to measure decision making as performance as there are scientific
investigations of it. A few examples are studies that operationalize decision making through
word memorization tasks, cognitive ability and problem solving tests, and tests of reading
comprehension. Currently, the primary focus for studies on decision making as performance has
been on teams. Not surprisingly, as seen in almost all team research, the type of task (e.g.,
Steiner, 1972) heavily influenced the decision making process. For disjunctive tasks, it was
found that the highest performing member of the team dominated the task (e.g., Bonner,
Baumann, & Dalal, 2002). For additive tasks that were more interdependent, teams consistently
outperformed individuals and were able to perform better the longer they stayed together
(Michaelsen, Watson, & Black, 1989; Vollrath, Sheppard, Hinsz, & Davis, 1989), becoming less
reliant on their best team member over time (Watson, Michaelsen, & Sharp, 1991).

7

Studies on cognitive decision making are few in number and have not gained significant
traction in the decision making literature. While the findings presented above are certainly
important for organizations, they are not consistent with main stream operationalizations of
decision making and, thus, better categorized in other areas of organizational psychology such as
individual and team performance. As a result, the remainder of this paper focuses on behavioral
decision making, the dominant approach to the study of decision making in the organizational
sciences as well as other scientific disciplines (e.g., economics, cognitive psychology).

Behavioral Decision Making
As was the case with cognitive decision making, the behavioral decision making
literature encompasses a wide variety of decision tasks. One type of task is strategic decision
making. Strategic decisions are primarily focused at the organizational level. The actual
decisions themselves are commonly ignored by aggregating to the level of the company. For a
decision to be included, it must be critical enough that it can affect organizational growth and
development. To date, little research exists examining this phenomenon. Initial evidence
suggests that strategic decisions are most effective when made through a thorough investigation
of all rational options, called procedural rationality, and made in the best interest of the corporate
political structure, called politics (Dean & Sharfman, 1995). Similarly, effective strategic
decisions are more likely made when companies have employees with a high level of
organizational loyalty and a high degree of competence (Dooley & Fryxell, 1999).
A second type of behavioral decision making is decisions with open-ended solutions.
This type of decision is similar to the cognitive decision making creativity tasks because
individuals are asked to provide an answer to a particular problem without being provided a set

8

of response options. However, it differs from the creativity tasks because the number of possible
solutions is not considered; rather, the single solution provided is evaluated for its quality or
consistency. The literature on open-ended decisions is primarily focused at the team level,
specifically interested in the relationship between leadership and effective team decision making.
It has been shown that teams are most effective when the team leader receives the problem from
the leader’s supervisor, interprets and frames the problem in a manner easily understood by team
members, and provides structure to how the team responds (Miner, 1979). Additionally,
involving subordinate team members in the decision process increases their positive affect
(Richter & Tjosvold, 1980).
Another type of behavioral decision making is called dynamic decision making. The goal
in studying dynamic decision making is to understand how individuals make decisions in a
highly complex dynamic environment, focusing on learning and incorporating feedback. The
tasks typically used to study this phenomenon simulate running a small company where money
or other firm resources need to be allocated to different organizational departments (e.g.,
inventory, sales, staffing). After making a series of decisions, the individual receives feedback
on the financial state of the company. The individual should then adjust his/her future decisions
based on the feedback to improve corporate performance. This process is then repeated for
several trials. With the goal of almost all of these studies to use feedback effectively to improve
performance, results have been rather mixed. While some researchers have found that different
modes of feedback lead to learning and increased performance (e.g., Atkins, Wood, & Rutgers,
2002), others have found that the complexity of the system overloads participants and, thus,
feedback is ignored (e.g., Diehl & Sternman, 1995; Gonzalez, 2005).

9

Along with evaluating the usefulness of feedback, some researchers have used dynamic
decision making to study the process of how individuals gain knowledge and formulate their
decisions within each trial. Research has shown that people tend to evaluate all information and
attempt to make rational choices (i.e., procedural rationality; Dean & Sharfman, 1995), even
when it is to their disadvantage to do so because of time pressure (Kleinmuntz & Thomas, 1987).
Additionally, it has been shown that some monetary choices in dynamic decision making trials
can be described using Tversky and Kahneman’s (1974) anchoring and adjusting heuristic.
Another paradigm that is very common in the organizational decision making literature is
discrete choice problems. These problems can be on a variety of organizational topics including
ethics (e.g., Detert, Trevino, Sweitzer, 2008), selection (e.g., Grove, 1981), marketing (e.g.,
Betsch, Haberstroh, Glockner, Haar, & Fiedler, 2001), and mergers and acquisitions (e.g., Pablo,
1994). Despite the wide range of topic areas, all discrete choice studies require participants to
choose one “best” option from a list of possible choices in a deterministic environment.
Research on discrete choice decisions is well established in the organizational sciences and has
yielded numerous results.
At the individual level, a number of common biases have been documented in the way in
which people make discrete choice decisions (Seale & Rapoport, 1997). Order of information
presented is very important (Blakeney & MacNaughton, 1971), with people focusing heavily on
early information (Dukerich & Nichols, 1991), and then seeking out additional information to
confirm their initial decisions (Betsch, et al., 2001). Despite this bias, individuals are commonly
unable to retrospectively determine which information or information gathering strategy they
used to make their decision (Viswesvaran & Barrick, 1992). Additionally, individuals tend to
use the majority rule heuristic (May, 1954; Russo & Dosher, 1983) to select an option when

10

dealing with multiple sources of information, even if the overall rating of the options is identical
(Zhang, Hsee, & Xiao, 2006). The impact of demographics on effective discrete decision
making is far less solidified with some evidence that demographics are not related to decision
effectiveness (Bridgeman & Marlowe, 1979) but other results suggesting that age and
intelligence were positively related to decision making effectiveness while risk-taking propensity
was negatively related to effective decision making (Taylor, 1975; Taylor & Dunnette, 1974).
Contextual factors also play a role in the discrete decision making process. For example,
type of task, organizational culture, and corporate politics have been shown to impact decision
effectiveness (Pablo, 1994). Similarly, when leaders provide advice to decision makers, leaders
who are more supportive are listened to more often, thus, resulting in higher overall performance
(O’Reilly, 1977). Over time, individuals who engage in procedural rationality by incorporating
all information and attempting to make a rational choice make more effective decisions than
those who make decisions based solely on prior success (Schweiger, Anderson, & Locke, 1985).
At the team level, communication is critical for discrete decision making success, similar
to findings for the creativity tasks in cognitive decision making. The more information flow
there is in a team, the more effective decisions the team makes (Holloman & Hendrick, 1972).
Similarly, in conditions of distributed expertise, it is vital that individuals within a team share
unique information (Stasser, 1992) so that team members will have greater team mental models
and transactive memory structures for making effective decisions (van Ginkel & van
Kippenberg, 2008, 2009). Finally, contrary to other types of decision making, stress was found
to improve team decision making by making team members more receptive to information from
one another (Driskell & Salas, 1991).

11

The final paradigm used to study behavioral decision making in the organizational
sciences is gambles. Gambles are a special case of discrete decision making where individuals
once again must select one option from a list of possible choices. Gambles, however,
incorporate a stochastic component by adding uncertainty to the outcomes with given or hidden
probabilities and a degree of risk associated with each possible choice. Each potential choice in
a gamble is called a prospect. Each prospect has multiple outcomes associated with it and each
outcome has a specific likelihood of occurrence represented by a probability ranging from zero
to 100 percent. Typically, within a given prospect, the probabilities of the outcomes sum to 100
percent. Gambles can be created using either monetary or nonmonetary outcomes; though, the
majority of gamble studies (even in the organizational sciences) utilize monetary outcomes
because much of the decision making literature using gambles comes from economics and other
financial fields. Despite the fact that most gamble studies are done utilizing lab tasks, evidence
exists that results generalize well to real world decisions (e.g., Wiseman & Levin, 1996).
Across monetary and nonmonetary gambles, many common relationships have been
discovered. At the individual level, one of the most consistent findings is that stress and affect
impact the riskiness of decisions (Andrade & Ariely, 2009; Ordonez & Benson, 1997; Seo &
Barrett, 2007). For example, individuals who exhibit signs of sadness tend to make riskier
decisions while individuals higher on anxiety tend to be more risk averse (Raghunathan & Pham,
1999). This is analogous to many risky decision making studies done in clinical psychology
using the Balloon Analogue Risk Task (BART) to assess an individual’s level of riskiness.
These studies frequently find that individuals with symptoms of depression (i.e., sadness) tend to
be more risk seeking (e.g., Lejuez, Aklin, Jones, et al., 2003; Lejuez, Aklin, Zvolensky, &
Pedulla, 2003; Lejuez et al., 2002; Lejuez, Simmons, Aklin, Daughters, & Dvir, 2004). In

12

general, feedback has been found to increase individuals’ ability to make effective decisions in
the presence of uncertainty over time (Zeelenberg & Beattie, 1997). However, fear of regret has
been found to ameliorate the effect of feedback on risky decisions such that individuals who fear
regretting past decisions ignore feedback and perform worse on subsequent decisions (Reb &
Connolly, 2009).
At the team level, the make-up of the team as well as the type of task has been shown to
impact how teams engage in making decisions under conditions of risk and uncertainty. It is
common for the riskiest member of the team to take the lead on tasks with a great deal of
uncertainty (van Kippenberg, van Kippenberg, & van Dijk, 2000). In teams working
individually as well as collectively, the effect of a loss experienced by a minority of team
members will not impact future risky decision making (Kameda & Davis, 1990). However, if
the majority of the team experiences a loss, future decision making of the team is impacted,
making the team less likely to engage in risky decisions (Paese, Bieser, & Tubbs, 1993; Tindale,
Sheffey, & Scott, 1993). Teams also tend to be more extreme than their individual members,
with the effect of a loss or gain shared by the entire team impacting future decisions to a greater
extent than if the loss or gain was experienced by any one individual (Whyte, 1993). Teams are
better able to make effective decisions in conditions of risk and uncertainty if they are good at
team communication, if minority opinions are expressed and discussed, if team leaders involve
all members in the final decision, and if individual team members engage in procedural
rationality (Colquitt, Hollenbeck, Ilgen, LePine, & Shepard, 2002; Hollenbeck, Ilgen, Sego,
Hedlund, Major, & Phillips, 1995; Hollenbeck, Colquitt, Ilgen, LePine, & Hedlund, 1998; Park
& DeShon, 2010). In contrast to other types of discrete decision making, when teams experience

13

stress and time pressure, information flow becomes restricted, impeding performance (Gladstein
& Reilly, 1985).

Limitations of Organizational Decision Making
While many interesting relationships between decision making and other processes have
been documented, a number of limitations exist. Perhaps the most prevalent shortcoming is the
unnatural way that most decision making is studied, making connections to the real world
extremely difficult. With the exception of gambles, all other decision making paradigms treat
the problem space as though the world is a perfectly deterministic closed system. By
guaranteeing decision outcomes, studies grossly misrepresent the complex, stochastic
environment in which employees operate. Creating a stochastic decision environment allows
researchers to understand how decision makers actually weigh potential alternatives with respect
to outcomes that are uncertain, mimicking real world problems.
In organizations, decision makers encounter a wide variety of problems that are
commonly influenced by multiple sources, a level of complexity that most decision making
studies ignore. Typically, even in stochastic decision making problems, participants answer a
single (simple) question and then immediately receive feedback on the result of their decision.
This once again lacks generalizability to the real world. Decisions in organizations rarely exist
in a vacuum; often times many decisions must be made before an employee is able to determine
the efficiency or effectiveness of their decisions. To represent this complexity, it would be
beneficial to have individuals make a series of small, similar decisions that will affect a larger
overall goal. This structure would better represent many common organizational problems like
those of the emergency medical teams and U.S. intelligence community mentioned earlier.

14

Similarly, organizational decisions rarely exist at only a single moment in time. Like
most topics in the organizational sciences, decision making studies suffer from an overreliance
on cross-sectional data collection and self-report methods. To better represent real workplace
decisions, it is necessary to collect data across multiple decisions over time to determine the
trends that individuals exhibit in their decision making and to detect any potential long run
biases. Likewise, tracking actual behavior rather than questionnaire responses to hypothetical
situations provides a more true to life representation of how employees would behave at work.
Overall, to best represent the stochastic, complex, dynamic nature of real organizational
decisions, researchers must employ problems with probabilistic (not guaranteed) outcomes,
many small decisions per trial related to an overarching goal, and have multiple trials to capture
the dynamics of how individuals change their decision making strategies over time.
Along with misrepresenting the complexities of real organizational decisions, there are
other shortcomings within the organizational decision making literature. One such problem is
that there is little research on common themes or utilizing common decision making tasks.
Unlike other areas such as clinical psychology which heavily rely on a common decision making
paradigm, organizational studies almost entirely utilize unique methods of researching the
decision making process. Compounding this problem is a lack of replication of findings within
or between decision making types. With the exception of stress, communication, and affect,
which have been studied in multiple decision making contexts, all other variables are unique to a
particular study or a particular decision type. This makes it very difficult to make strong,
confident recommendations about what organizational variables affect the decision making
process and how that process affects important organizational outcomes. Not surprisingly, then,

15

an additional limitation is that there are few tangible recommendations for organizations on how
to impact decision making in ways that will likely increase organizational outcomes of interest.
The final major shortcoming with the decision making literature in the organizational
sciences is the lack of attention paid to general decision making models that can describe the
decision process. In other areas of study such as cognitive psychology and economics, much
attention is paid to creating general models of decision making that can accurately capture the
biases that individuals engage in when making certain types of decisions. Along with little
theory on general decision making models, there is also rather limited attention paid to the
descriptive decision models used in other scientific disciplines. While across different scientific
fields the content of problems will certainly change, it is entirely possible that the underlying
decision making process will remain constant. Therefore, it is important to incorporate decision
models from related scientific fields and test their predictions and findings within an
organizational context to better understand the process by which individuals make decisions.
This will provide insight into potential systematic biases that decision makers exhibit in a variety
of contexts, allowing researchers to potentially create training interventions to help employees
make more optimal decisions.

Prospect Theory: Describing Risky Decision Making
Unlike decision making research in the organizational sciences, other fields frequently
create mathematical models in an attempt to describe different decision making types. Gambles,
in particular, have received significant attention in cognitive psychology and economics.
Initially, individuals were modeled as purely rational beings that would always choose the best
possible prospect despite the type of question asked. Under these assumptions, Expected Utility

16

Theory (Atkinson, 1957; Keeney & Raiffa, 1976) was the dominant rational choice model of
human behavior. However, under conditions of risk and uncertainty, individuals were found to
commonly deviate from making rational choices. As a result, Kanheman and Tversky (1979)
created Prospect Theory in an attempt to explain some common biases in the choices people
made when confronted with risky decisions. Prospect Theory was a significant improvement
over Expected Utility Theory and other rational choice models, but was quite limited and still
failed to accurately describe risky decision making over a sequence of decisions, could only
incorporate a small number of prospects, and was found to inadequately account for the effect of
question framing. Therefore, Tversky and Kahneman (1992) created a more general theory of
risky decision making entitled Cumulative Prospect Theory. Additionally, Thaler and Johnson
(1990) and Staw (1976, 1981) created extensions of Prospect Theory for how individuals make
decisions over a sequence of decisions called the house money effect and escalation of
commitment, respectively. Cumulative Prospect Theory and its generalizations have been shown
to accurately describe many aspects of the risky decision making process and as a result are
currently the dominant approach to the study of gambles. To understand how prospect theory
and its generalizations could be used to predict risky decision making during information
acquisition, each theory will be briefly explained, highlighting its origin, key features, and
primary predictions.

Prospect Theory
Individuals exhibit many common tendencies when forced to make a decision that has
multiple probabilistic outcomes (i.e., gambles). For example, one common bias in risky decision
making is called the certainty effect (Allais, 1953). When individuals are forced to choose

17

between a set of risky prospects and an equal probability outcome is removed from each
prospect, the psychological effect of removing the probability is strongest for prospects that
originally had certain (probability = 0% or 100%) outcomes. As a result, even though the
ordering of prospect preference should not change by removing an equal outcome from all
choices, individuals tend to switch preferences from a once certain outcome to one that was
originally uncertain despite the fact that the once certain outcome is still objectively better
(MacCrimmon & Larson, 1979; Slovic & Tversky, 1974). Additionally, when individuals are
forced to choose between a set of high probability prospects with positive outcomes, the vast
majority of people will select the most certain option even if it has a lower possible payout.
Conversely, when individuals are forced to choose between low probability prospects, they will
frequently choose the riskier option with the highest possible payoff (Kahneman & Tversky,
1979).
It turns out that the exact nature of the certainty effect is more complex and also depends
on whether the prospects refer to positive or negative outcomes. When individuals are forced to
choose between a set of prospects with negative outcomes, preferences systematically switch.
For high probability prospects, individuals will choose the option with less certainty, in an
attempt to avoid any loss. For low probability prospects, individuals will choose the outcome
with more certainty, in an attempt to avoid the possibility of a really large loss. This preference
reversal is called the reflection effect (Fishburn & Kochenberger, 1979; Markowitz, 1952;
Williams, 1966). Taken together, the certainty and reflection effects create a pattern of behavior
when deciding among risky prospects; outcomes with certainty are systematically over weighted,
resulting in individuals being risk averse with positive outcomes and risk seeking with negative
outcomes (Kahneman & Tversky, 1979). This is particularly important because these findings

18

violate one of the key components of Expected Utility Theory, that individuals are always risk
averse.
The final common bias that individuals demonstrate when making risky decisions is
called the isolation effect (Kahneman & Tversky, 1979). When evaluating risky prospects,
individuals tend to ignore aspects that the prospects have in common. This can be problematic
because there is frequently more than one way to decompose a problem, potentially resulting in
different preferences. More importantly, these findings suggest that decisions are not made
solely by the anticipated final state that the individual will likely end up in. Rather, that
decisions are made based on an anticipated change from the current state, or reference point,
with which the individual begins the problem. These findings once again demonstrate that
individuals violate the tenants of Expected Utility Theory in the presence of risky decisions
because in Expected Utility Theory end states (not changes) are solely the basis for the selection
of prospects.
The systematic biases that decision makers exhibit in the presence of risk demonstrates
that Expected Utility Theory and other rational choice models of decision making inadequately
describe human behavior. As a result, Kahneman and Tversky (1979) built an alternative
mathematical representation of risky decision making, called Prospect Theory (PT), in the hope
of accounting for biases like the certainty, reflection, and isolation effects. PT was originally
created to describe how individuals make a single risky decision with two or three prospects and
given probabilities. However, Kahneman and Tversky note that the model is generalizable to
alternative, more complex decision types.
PT models the decision process as the result of two phases, the editing phase and the
evaluation phase. The editing phase suggests that individuals psychologically frame and edit

19

each potential prospect to simplify the decision making process through a series of rules prior to
selecting the most desirable option. The first rule is called Coding, where individuals code each
prospect in terms of potential gains and losses relative to a reference point. The second rule is
called Combination, which states that individuals combine probabilities that are related to
identical outcomes. The next rule is labeled Segregation, which is where individuals separate
components of prospects with certainty from components with uncertainty. The final editing rule
is Cancellation, when individuals will cancel shared components across prospects. The result of
going through each rule of the editing phase is that each prospect has been reframed and
simplified so that individuals can more easily make a decision as to the most desirable prospect.
At the conclusion of the editing phase, individuals begin the evaluation process of the
newly framed prospects. The evaluation phase of PT introduces two new components of the
each possible outcome in a given prospect is assigned a value, 𝜐, which represents the subjective
risky decision making process that prior rational choice models did not include. The first is that

value of the outcome. Since outcomes are thought to be considered relative to a reference point,
the value function measures the outcome in deviations (positive or negative) from that reference
point. This suggests that the subjective representation of the prospect may not accurately reflect
the stated problem.
each prospect is weighted by some weighting function, 𝜋. The weighting function indicates the
The second new aspect of the risky decision making process is that the probability of

impact of a given outcome probability on the overall probability of choosing the prospect. An
important feature of the weighting function is that frequently, the weights within a given

prospect do not sum to one, despite the fact that the probabilities they are weighting will sum to

20

one. This allows for the common finding that individuals’ psychological representation of
probability does not perfectly match (a deterministic) reality (Edwards, 1962; Fellner, 1961).
Prospect Theory is designed to handle three specific types of decision making gambles.
The first type is where all possible outcomes in all prospects are positive, called the strictly
positive condition. The second is where all possible outcomes in all prospects are negative,
called the strictly negative condition. Finally, the third condition is where both positive and
negative outcomes are in each prospect, called the normal condition (Edwards, 1962; Markowitz,
1952). Normal gambles are the most general and most studied type of risky decision making
process and, thus, will be the focus of subsequent sections. In normal gambles, prospects are
example, if a prospect has two possible outcomes, 𝑋 and 𝑌, then the prospect would be normal if
considered regular if they have an equal number of outcomes above and below zero. For

𝑋 ≤ 0 ≤ 𝑌 or if 𝑋 ≥ 0 ≥ 𝑌. For regular prospects in normal gambles, the value function, υ,

and the weighting function, π, are combined with each outcome and probability, respectively,
within a prospect and then summed to give the overall value of the prospect. Equation 1
𝑉(𝑥, 𝑝; 𝑦, 𝑞) = 𝜋(𝑝)𝜐(𝑥) + 𝜋(𝑞)𝜐(𝑦),

demonstrates this mathematically:

where 𝑉 is the overall value of the prospect, 𝑥 is the first possible outcome, 𝑝 is the probability

(1)

associated with outcome 𝑥, 𝑦 is the second possible outcome, 𝑞 is the probability associated with
outcome 𝑦, 𝜐 is the value function for the prospect, and 𝜋 is the weighting function for the
prospect.

Both the value function and the weighting function have been found to have consistent
qualities. As previously mentioned, the value function is designed to represent the magnitude of
a given outcome in terms of deviations from some reference point. This is designed to replicate

21

findings in the physical sciences that show that people frequently are poor judges of absolute
values of physical stimuli like sound, light, and temperature. Rather, people are significantly
better at judging the change in these states from some reference point (e.g., Helson, 1964). For
both positive and negative outcomes, individuals weight the value of a deviation from the
reference point larger, the closer it is to the reference point. For example, the difference between
a loss of $10 and $20 is weighted with more impact than an equally large difference between a
loss of $1000 and $1010. As a result, the slope of the value function decreases as it moves
further away from the reference point. This leads to the common finding in risky decision
making that the shape of the value function is concave above the reference point and convex
below it (Galanter & Pliner, 1974; Kahneman & Tversky, 1979). Additionally, it is found that
individuals weight negative outcomes greater than they do positive outcomes as indicated by the
certainty and reflection effects mentioned earlier. As a result, the slope of the value function is
greater below the reference point than it is above it (e.g., Barnes & Reinmuth, 1976; Grayson,
1960; Green, 1963; Halter & Dean, 1971; Swalm, 1966). Figure 1 shows a hypothetical value
function for a normal gamble with regular prospects. The left side the figure shows the value
function for negative outcomes and the right side of the figure shows the value function for
positive outcomes. It can be seen that the slope for negative outcomes is steeper than the slope
for positive outcomes and that for both types of outcomes, the slopes decrease with increased
distance from the reference point.

22

Value

Losses / Gains

Figure 1. Hypothetical value function for Prospect Theory.

Like the value function, the weighting function has a number of unique properties.
Although the weighting function is related to outcome probabilities, it is not considered a
measure of the belief that a particular outcome will happen. This means that it is not a measure
of perceived likelihood of an outcome but rather, it measures the perceived impact of the
possible outcome on the overall desirability of a prospect. One common finding is that very low
probabilities are over weighted (Kahneman & Tversky, 1979). It is important to note that the
over weighting of a probability is distinct from the overestimation of the likelihood of that
probability (see Kahneman & Tversky, 1973). Because the weighting function is a measure of
impact, the overweighting of low probabilities means that individuals weight the impact of their
happening greater than they should. This effect has the potential to interact with the
23

overestimation of the likelihood of low probability events to seriously alter decision making
behavior for very rare instances. Otherwise, individuals are found to underweight the impact of
moderate to high probable events, which is why the total summed weighting function for a given
prospect is likely less than one, despite the fact that the actual probabilities in the prospect sum to
one. Figure 2 shows the hypothetical weighting function for a normal gamble with regular
prospects. The dotted line represents the ideal situation where an individual perfectly weights
the impact of the probability equal to the actual probability. Meanwhile, the solid line represents

0.8
0.6
0.0

0.2

0.4

Decision Weight: Pi(p)

1.0

the proposed weighting function based on the biases predicted by PT.

0.0

0.2

0.4

0.6

0.8

1.0

Stated Probabiliy: p

Figure 2. Hypothetical weighting function for Prospect Theory.
Overall, PT significantly advanced understanding of the decision making process under
conditions of risk. The addition of an editing phase as well as the value function and weighting
24

function in the evaluation phase allowed researchers a tool with which to model many of the
common biases found in risky decision making such as the certainty, reflection, and isolation
effects. As a result, it allowed for a better overall description of the process by which individuals
decide which prospect is most desirable when making a gamble.

Cumulative Prospect Theory
After the creation of the original Prospect Theory, a number of limitations were
discovered. First, Prospect Theory only can handle very simple gambles with two or three
possible prospects. Also, the outcomes in each prospect must have stated probabilities, so
Prospect Theory cannot properly predict decision making under situations of ambiguity (i.e.,
individuals are not told what the outcome probabilities are). Additionally, the strength of the
effect of question framing was better realized. It became apparent that the way in which a
question is worded can significantly affect the biases that individuals demonstrate in risky
decision making (Tversky & Kahneman, 1986). Tversky and Kahneman also discovered that the
effect that losses loom larger than gains was more significant than originally believed
(Kahneman & Tversky, 1984; Tversky & Kahneman, 1991). As a result, many authors began to
generalize Prospect Theory to include an infinite number of prospects, have different weighting
functions for gains and losses, and handle situations of both risk and ambiguity (Quiggin, 1982;
Schneidler, 1989; Weymark, 1981; Yaari, 1987). Therefore, Tversky and Kahneman (1992)
built upon this work and created a more general form of Prospect Theory known as Cumulative
Prospect Theory (CPT).
CPT is very similar to the original Prospect Theory in a number of ways. It once again
has two phases, editing and evaluation, and can handle three types of gambles: all positive, all

25

negative, or mixed (called normal in Prospect Theory). For mixed gambles, the overall value of
a prospect is once again the summation of each outcome weighted by a value function and each
probability weighted by a weighting function. However, the difference is that there are separate
weighting functions for both positive and negative outcomes. Shown in Equation 2, the overall

𝑓

+

value of a prospect can be mathematically represented as:
𝑉(

𝑓

+

𝑉(𝑓) = 𝑉(

𝑓

−

𝑓

−

)

−
0
𝑛
) = ∑ 𝑖=0 𝜋+ 𝜐( 𝑥 𝑖 ), 𝑉( 𝑓 ) = ∑ 𝑖=−𝑚
𝑖

𝑉(𝑓) is the overall value of the prospect, 𝑉(
𝑉(

) + 𝑉(

𝑓

+

𝜋 − υ( 𝑥 𝑖 ).
𝑖

(2)

) if the overall value of all positive outcomes,

) is the overall value of all negative outcomes, 𝑛 is the number of total positive outcomes,

𝑚 is the number of total negative outcomes,

outcome, 𝑖,

𝜋 + is the weighting function for each positive
𝑖

𝜋 − is the weighting function for each negative outcome, 𝑖,
𝑖

function for each outcome, and

𝑥𝑖

𝜐 is the overall value

represents each outcome.

Equation 2 demonstrates that the overall value of each prospect is the sum of all positive
outcomes weighted by the overall value function and the individual outcome’s probability and
positive weighting function combined with the sum of all negative outcomes weighted by the
overall value function and the individual outcome’s probability and negative weighting function.
Additionally, Equation 2 has the appealing mathematical property of having both the positive
and negative weighting functions sum to one, without loss of generalizability.
Splitting the weighting function allows researchers to more accurately model the biases
detected in risky decision making. One such finding that can be accurately modeled by CPT is
that individuals are risk seeking for gains and risk averse to losses of low probability outcomes

26

while being risk averse to gains and risk seeking for losses for high probability outcomes
(Kahneman & Tversky, 1979; Tversky & Kahneman, 1992). The shape of the weighting
functions is also different than in the original Prospect Theory, to be more consistent with
empirical results (see Camerer, 1992 and Camerer & Ho, 1994 for reviews). Figure 3 shows a
hypothetical weighting function in CPT. As in Prospect Theory, low probability outcomes are
generally over weighted and moderate probability outcomes are generally underweighted.
However, the weighting functions now more accurately predict that individuals accurately
weight extremely low and high probability events so at the extremes the ideal (dotted) and

0.0

0.2

0.4

w(p)

0.6

0.8

1.0

hypothesized (solid line) functions are much closer.

0.0

0.2

0.4

0.6

0.8

1.0

p

Figure 3. Hypothetical weighting function for Cumulative Prospect Theory.

27

The benefits of CPT above and beyond prospect theory are important when modeling
complex decisions with many possible alternatives. Splitting up the weighting function for
positive and negative outcomes as well as the ability to model an infinite number of prospects
and outcomes allow researchers to use Cumulative Prospect Theory to study many different
types of risky decision making. As a result, CPT is the dominant approach to modeling and
studying decision making under conditions of risk.

The House Money Effect and Escalation of Commitment
While Cumulative Prospect Theory does an excellent job in modeling the decision
making process of individuals confronted with situations of risk, it is designed to study only
single decisions. Frequently in organizations, it is necessary for individuals to make a series of
decisions over time toward the solution of a larger goal. It is likely that the results from prior
decisions have an impact on how individuals make subsequent choices (Arkes & Blumer, 1985;
Thaler, 1980). Research on the house money effect (Thaler & Johnson, 1990) and escalation of
commitment (Staw, 1976, 1981) investigate how individuals update their risky decision making
process based on successes and failures to prior risky decisions.
Thaler and Johnson (1990) extend Prospect Theory by the addition of a new editing rule
called Quasi-hedonic editing, which suggests that individuals will edit potential prospects in the
way that makes the individual happiest. This is supported by evidence that individuals will
integrate or segregate prior positive and negative outcomes from subsequent decisions based on
what is most psychologically pleasing. For example, people tend to prefer to spread out
experiencing positive outcomes to separate the resulting happiness. Additionally, large losses
desensitize an individual to subsequent smaller losses while small losses sensitize individuals to

28

potential larger losses. As a result, small losses after a much larger one seem less disturbing and
after a small loss individuals alter their behavior in an attempt to avoid any more unpleasant
outcomes. Also, a small loss after a prior gain is immediately integrated into the gain, resulting
in the loss having a minimal negative effect on the individual. All of these findings imply that
individuals do not simply integrate prior outcomes into their reference point, as Prospect Theory
would suggest, but rather they are separated and can directly impact subsequent decisions.
Perhaps the most common result of this process is what is termed the house money effect.
After experiencing a positive outcome, individuals will become risk seeking in subsequent
decisions, because they are essentially gambling with “house money.” Combined with the
finding that individuals become risk averse after experiencing a negative outcome, these findings
are in direct contradiction to Prospect Theory and Cumulative Prospect Theory which have
shown individuals to be risk averse for potential positive outcomes and risk seeking for negative
outcomes. The house money effect has been quite robust and the pattern of results is the same
for both real and hypothetical gambles, implying that the risky decision making process changes
over time based on previous experience (Thaler & Johnson, 1990).
While the house money effect has been consistently demonstrated for positive outcomes,
the evidence is more complex when dealing with prior negative outcomes. In fact, individuals
have been shown to be incredibly risk seeking even after a series of very negative outcomes,
directly contradicting the findings of the house money effect for negative outcomes. This
phenomenon is called escalation of commitment (Laughhunn & Payne, 1984; Staw, 1976, 1981).
When individuals initially engage in a risky behavior that leads to negative outcomes,
they may continue to engage in that behavior if the decision has the potential to lead to an
eventual positive result, making up for all prior negative effects. While it was initially thought

29

that individuals would only escalate their commitment if they chose the initial risky behavior
(Staw, 1976), it was later found that even when forced to engage in the initial risk, some people
will continue to invest resources if the potential is there to break even (Staw, 1981). Many
possible explanations for this phenomenon have been explored, including that individuals may
bias their view of the initial decision to be more consistent with the negative outcome, thus
reducing cognitive dissonance and justifying their continued course of action (Aronson, 1968,
1972; Festinger, 1957). Additionally, it has been thought that individuals may engage in
continued risky behavior due to the gambler’s fallacy that the continued investment is bound to
pay off eventually because they are “due for a win” (Lee, 1971; Staw, 1976).
Taken together, findings from the house money effect and escalation of commitment
demonstrate that the risky decision making process changes when dealing with a dynamic
environment in which individuals engage in multiple decisions and receive feedback of the
results of those decisions. As a result, it is necessary to incorporate findings from these theories
as well as from Prospect Theory and Cumulative Prospect Theory to fully understand and
describe the risky decision making process over time.

Risky Decision Making during Information Acquisition
In organizations it is common for individuals to have to make decisions in content areas
where they are not experts. As a result, they must first acquire relevant information before being
able to deduce an appropriate solution. Unfortunately, like the intelligence officers and
emergency medical staff described earlier, it can be unclear as to what information is most
critical and most relevant. Therefore, when individuals engage in searching for new information,
they are essentially engaging in a sequence of gambles with each decision to attend to a new

30

piece of information being a unique gamble. Each of these gambles will likely have two
potential prospects. The first prospect is that the individual chooses to attend to the piece of
information. This prospect has two distinct outcomes, that the information was relevant and
aided in the solution of the problem or that it was irrelevant and either simply wasted time or,
even worse, provided misinformation that can bias the problem’s solution. The second prospect
is that the individual can ignore the piece of information. This prospect once again has two
potential outcomes, that the individual correctly ignored an irrelevant piece of information, thus,
not wasting time or acquiring misinformation or that the information was actually relevant and
the individual will have insufficient knowledge for solving the problem, likely leading to an
inefficient or ineffective solution.
Given the criticality of individuals being able to effectively make decisions for
organizational success (e.g., Shrode & Brown, 1970; Singh, 1986; Ziller, 1957), it is imperative
that researchers understand the process by which individuals choose what information to attend
to and what to ignore. To date, no research exists that directly studies this important
phenomenon. While the organizational decision making literature does contain important
findings for understanding the process of decision making under risk and uncertainty, it lacks a
general framework to describe the decision making process. Prospect Theory (PT) and its
generalizations have been found to be predictive of the biases that individuals engage in when
making risky financial decisions in a wide variety of organizational contexts such as firm
resource allocation (e.g., Bromiley, 2009), organizational strategy (e.g., Bromiley, 2010), and
performance testing (e.g., Bereby-Meyer, Meyer, & Budescu, 2003; Bereby-Meyer, Meyer, &
Flascher, 2002). Additionally, time and information have been repeatedly argued to be valuable
resources synonymous with monetary outcomes, which is the original outcome described by PT

31

(Ariely & Zakay, 2001; Cunah & Caldero, 2009; Katsikopoulos, Duse-Anthony, Fisher, &
Duffy, 2002; Leclerc, Schmitt, & Dube, 1995). Therefore, PT provides a good foundation for
understanding the process of making risky decisions while acquiring information in
organizations.
The purpose of the current research is to understand the decision making process that
individuals engage in when acquiring new information. In an attempt to fill the gaps identified
in the organizational decision making literature, the study will integrate findings from
organizational studies on gambles with the predictions of PT and its generalizations and common
relationships found in the PT literature. Doing so will allow for understanding what general
biases individuals engage in when making risky decisions related to information acquisition, a
phenomenon that has previously not received attention in the literature. Additionally, using
previous findings in the Prospect Theory literature as well as organizational studies with gambles
provides insight into what individual differences play a role in impacting the risky decision
making process. In addition, this study aims to understand how the decision making process
changes over time and, thus, will explore how individuals make a sequence of small decisions
related to a series of larger goals. This is important because it more closely represents the types
of decision making problems that employees face in organizations.
Utilizing Prospect Theory and its generalizations to explain the decision making process
individuals engage in when searching for new information has many potential benefits. It
provides insight into the potential biases that people could make while deciding what
information to attend to and what to ignore. Discovering these biases will inform researchers as
to the types of training interventions necessary to help decision makers perform more efficiently
and more effectively, likely resulting in higher overall performance. Similarly, understanding

32

how this process unfolds over time will provide evidence about how feedback and prior
decisions are incorporated and alter the decision making process. This once again has
implications for how results from decision making tasks should be structured to help employees
learn what are the most effective decision making strategies when acquiring new information.
Additionally, testing the predictions of prospect theory for the decisions necessary for
information acquisition will also inform potential generalizations or boundary conditions of
Prospect Theory. Since the vast majority of PT studies have been done with quantifiable
outcomes (i.e., money and time), it is currently unknown whether the theory can describe the
decision making process for prospects with more subtle successes and failures, like those
encountered during information acquisition. Likewise, most studies on gambles use hypothetical
questions rather than real behavior for understanding risky decision making. Using real
behaviors will provide additional information to researchers as to the ability of PT to describe
how individuals actually behave. Finally, few studies exist that thoroughly test the dynamics of
the decision making process; rather, most studies that look at prior performance utilize only two
or three time points (e.g., Staw, 1976, 1981; Thaler & Johnson, 1990). Having individuals
engage in a long sequence of decisions will provide further evidence for how feedback and prior
performance affect the biases commonly described by PT and its generalizations.

Prospect Theory Hypotheses
Despite the widespread use and acceptance of PT, it has never been used to describe
decision making during information acquisition, with usefulness of information as an outcome.
As a result, it is first necessary to determine if the common biases in decision making that PT
describes are even present when making gambles about acquiring information. PT predicts

33

specific patterns of behavior based on each of its components for individuals making gambles.
The first behavioral prediction is based on the shape of the weighting function. The way in
which a question is framed dramatically impacts behavioral decision making (Kahneman &
Tversky, 1986; Kuhberger, 1998, van Schie & van der Pligt, 1995). As such, Cumulative
Prospect Theory divided the weighting function into two separate components, one for positive
outcomes and one for negative outcomes (Tversky & Kahneman, 1992). By having two separate
functions it is possible to model the riskiness of positive and negative outcomes separately for
high and low probability events. It has been found that individuals tend to be risk seeking for
positive outcomes with low probabilities and for negative outcomes with high probabilities.
Conversely, individuals tend to be risk averse for negative outcomes with low probabilities and
positive outcomes with high probabilities (Abdellaoui, Vossman, & Weber, 2005; Camerer &
Ho, 1994; Etchart-Vincent, 2004; Gonzalez & Wu, 1999). Therefore, it is probable that when
individuals are presented with information framed positively with a high likelihood of being
relevant, they will be risk averse.
Hypothesis 1a. Individuals will be risk averse when faced with positively framed
information with a high likelihood of relevance.
Alternatively, it is likely that individuals will be risk seeking when they are presented with
negatively framed information with a high likelihood of being relevant.
Hypothesis 1b. Individuals will be risk seeking when faced with negatively framed
information with a high likelihood of relevance.
In addition, when individuals are presented with low probability information, they will most
likely be risk seeking if the information is framed positively and risk averse when the
information is framed negatively.

34

Hypothesis 1c. Individuals will be risk averse when faced with negatively framed
information with a low likelihood of relevance.
Hypothesis 1d. Individuals will be risk seeking when faced with positively framed
information with a low likelihood of relevance.
The shape of the value function in Cumulative Prospect Theory (Figure 1) implies that
people value negative outcomes significantly more than positive outcomes, meaning that the
effects of riskiness for negative outcomes will be larger than the effects for positive outcomes
(Weber, 1994).
Hypothesis 2. The effects in hypotheses 1b and 1c will be greater than the effects in
hypotheses 1a and 1d.
The shape of the weighting function in Cumulative Prospect Theory (see Figure 3)
indicates that individuals have a tendency to overweight information that has a very low
likelihood of being relevant, underweight information that has a moderate likelihood of being
relevant, and approximately accurately weight information with a very high likelihood of being
relevant (Bruhin, Fehr-Duda, & Epper, 2010). Additionally, individuals have a tendency to
overestimate the likelihood of rare events, making them appear more likely than they really are
(Kahneman & Tversky, 1973). These findings imply that individuals should attend to
significantly more information that has a very low likelihood of relevance than is expected by the
actual probability of relevance. Alternatively, individuals should attend to less information with
a moderate likelihood of relevance than is expected by the actual probability of relevance.
Finally, individuals should attend to the correct amount of information that has a high likelihood
of relevance.

35

Hypothesis 3a. Individuals will attend to significantly more information of low relevance
than the actual probability of relevance.
Hypothesis 3b. Individuals will attend to less information of moderate relevance than the
actual probability of relevance.
Hypothesis 3c. Individuals will attend to an approximately equal amount of information
of high relevance compared to the actual probability of relevance.
The prior predictions are designed to describe single decisions or general tendencies and
do not account for changes in behavior over time. That said, it is commonly demonstrated that
prior successes and failures can drastically alter the pattern of biases in risky decision making
(Ansic & Keasey, 1994; Pietras, Locey, & Hackenberg, 2003; Staw, 1976, 1981; Thaler &
Johnson, 1990). In particular, the house money effect has received significant support in the
literature. It is commonly found that individuals are risk seeking for all types of outcomes after
experiencing a prior success. Similarly, individuals have also been found to be more risk averse
after experiencing a prior failure (Ackert, Charupat, Church, & Deaves, 2006; Akers, Joyner,
Pezzo, Nash, Siegal-Jacobs, & Stone, 1994; Cowley, 2008; Hollenbeck, Ilgen, Phillips, &
Hedlund, 1994). This finding contradicts the predictions of the escalation of commitment theory.
It is possible that the escalation of commitment effect needs extremely specific circumstances to
be witnessed, circumstances that are not often present. For example, it has been found that for
escalation of commitment to occur, the decision must be one in which it is extremely difficult to
change courses of action (Brehm & Cohen, 1962). In most gamble research this is not the case,
which could explain why the predictions of escalation of commitment are commonly
unsubstantiated (e.g., Weber & Zuchel, 2005; Zeelenberg & van Dijk, 1997). Therefore, it is
likely that there will be a positive effect between prior outcomes and riskiness on subsequent

36

decisions. When individuals experience a prior success, they will make more risky decisions.
Alternatively, when individuals experience a prior failure, they will engage in less risky
decisions.
Hypothesis 4a. After a prior positive outcome, individuals will make more risky
decisions.
Hypothesis 4b. After a prior negative outcome, individuals will make less risky decisions.

Individual Differences Hypotheses
The literature on prospect theory as well as many organizational studies of risky decision
making investigate how individual differences affect the riskiness and effectiveness of decisions.
The relationship that has received the most attention in the literature is between affect and
decision riskiness. However, affect is rarely examined at the trait level and has not been tested
as to whether it has a main effect across time. Frequently, affect will be manipulated in a lab
setting to determine its effect on the riskiness of decisions. Affect has been found to alter the
weighting function in cumulative prospect theory, resulting in a change in the nature of biases
(Rottenstreich & Hsee, 2001; Seo, Goldfarb, & Barrett, 2010). More specifically, it has been
found that people who experience greater negative affectivity are more prone to make risky
decisions (Raghunathan & Pham, 1999), and more likely to make less effective decisions
(Taylor, 1975). Additionally, individuals who have experienced a positive emotion will be more
likely to overestimate the probabilities of positively framed prospects and be more conservative
overall, resulting in significantly more risk averse behavior (Nygren, Isen, Taylor, & Dulin,
1996). Thus, individuals who are higher on negative affectivity should make more risky decision
as well as perform worse overall relative to those lower on negative affectivity. Conversely,

37

individuals who are higher on positive affectivity should overweight positively framed prospects
and be more risk averse relative to those who are lower on positive affectivity.
Hypothesis 5a. Individuals higher on negative affectivity will make riskier decisions
relative to those lower in negative affectivity.
Hypothesis 5b. Individuals higher on negative affectivity will perform worse overall
relative to those lower in negative affectivity.
Hypothesis 5c. Individuals higher on positive affectivity will choose to attend to more
positively framed information compared to those lower on positive affectivity.
Hypothesis 5d. Individuals higher on positive affectivity will make less risky decisions
compared to those lower on positive affectivity.
Another individual difference that has received significant attention is the effect of
motivation and goals on decision making. Like affect, the effect of goal orientation as an
individual difference is rarely considered and it is currently unknown as to the effect that goal
orientation has across time. Goals are frequently manipulated in a lab setting to determine their
effect on the riskiness and effectiveness of decisions. Not surprisingly, they have been found to
be very important to decision making of both individuals (e.g., Heath, Larrick, & Wu, 1998) and
teams (e.g, Lepine, Hollenbeck, Ilgen, & Hedlund, 1997). Specifically, regulatory focus
(Higgins, 1997, 2000) has been found to drastically impact the pattern of risky decision making.
In regulatory focus theory, individuals are considered to be either in a state of prevention or a
state of promotion. The prevention state is where individuals do just enough to not fail.
Alternatively, a promotion state is where individuals aim to exceed expectations. Individuals in
a prevention state tend to be significantly more averse to negative outcomes, making them take
fewer risks in decision making. Conversely, individuals in a promotion state tend to be more risk

38

taking in hopes of achieving greater positive outcomes (Halamish, Liberman, Higgins, & Idson,
2008; Kluger, Stephan, Ganzach, & Hershkovitz, 2004; Scholer, Zou, Fujita, Stroessner, &
Higgins, 2010). The concepts of prevention and promotion states of motivation in the regulatory
focus literature are very similar to performance avoid and performance approach goal orientation
in the organizational literature on motivation (Bandura, 1997; Locke & Latham, 2002). As such,
it is likely that individuals who have a performance avoid goal orientation will be more loss
averse and, thus, make fewer risky decisions. Individuals with a performance approach goal
orientation, on the other hand, should take more risks in hopes of achieving greater success.
Hypothesis 6a. Individuals higher on performance avoid goal orientation will make
fewer risky choices compared to individuals lower on performance avoid goal
orientation.
Hypothesis 6b. Individuals higher on performance approach goal orientation will make
more risky choices compared to individuals lower on performance approach goal
orientation.
The final two individual differences that have been researched in relation to risky
decision making are risk taking propensity and cognitive ability. The body of literature on these
two constructs related to risky decision making is significantly less, however a few important
relationships have been discovered. Not surprisingly, individuals who report a greater propensity
for risk taking have been shown to take more actual risks (Taylor & Dunnette, 1974). However,
this relationship tends to weaken when actual decisions are studied across time (e.g., Sitkin &
Weingart, 1995). Therefore, individuals who report themselves as higher on risk taking
propensity should take slightly more risks when acquiring information over time.

39

Hypothesis 7. Individuals higher on risk taking propensity will make more risky choices
compared to those lower on risk taking propensity.
Cognitive ability has also been shown to impact risky decision making at both the individual
(e.g., Taylor & Dunnette, 1974) and team (e.g., Lepine, 2003, 2005) levels, with those higher on
cognitive ability taking fewer risks and performing better overall on risky decision making tasks.
As a result, those higher on cognitive ability should take fewer risks and perform better.
Hypothesis 8a. Individuals higher on cognitive ability will take fewer risks compared to
those lower on cognitive ability.
Hypothesis 8b. Individuals higher on cognitive ability will perform better than those
lower on cognitive ability.

Learning and Decision Making Hypotheses
Despite the focus of a wealth of research on sequential risky decision making, very few
studies have actually looked at the effect of learning. While findings such as the house money
effect and escalation of commitment suggest that individuals do alter decision making strategies
based on prior outcomes, they do not answer the question as to whether individuals learn or get
better at making risky decisions over time. One potential reason for this is that most studies
looking at sequential decision making utilize very few (T < 5) time points. That said, there is
some initial evidence that suggests that as individuals make a long series of risky decisions, their
choices begin to mirror the predictions of rational choice models (Aloysius, 2007), indicating
that they learn to effectively navigate an uncertain environment over time. In addition, there is
evidence that individuals can learn unstated probabilities over time and alter their decision
making to better represent underlying probabilities (Pleskac, 2008; van de Kuilen & Wakker,

40

2006). This suggests that as individuals engage in making a series of decisions related to
information acquisition they will improve over time. Also, these findings imply that individuals
weighting function for probabilities will change over time such that the more decisions that are
made, the closer individuals’ decision weights will mirror the ideal rather than predicted
weighting function from Prospect Theory (dashed line in Figure 2).
Hypothesis 9a. Individuals will perform better over time.
Hypothesis 9b. Over time, individuals will attend to information more closely to the
proportion of time indicated by the true probability.
Finally, recent evidence exists that when individuals make a sequence of very similar
decisions, eventually the decisions become automated. As a result, learning effectively stops
after a certain point and the decision maker will always choose the same option from a given
prompt (Pleskac & Wershbale, 2011). Given that individuals make a large number of risky
decisions related to information acquisition, it is likely at some point the process will become
automated and decision making time will significantly decrease and all decisions about similar
prospects will become the same.
Hypothesis 10a. Over time, individuals will make decisions about similar prospects
quicker.
Hypothesis 10b. Over time, individuals will make the same decision for all identical
prospects.

41

METHOD

Sample
Data were collected from 245 undergraduate psychology students from a large
Midwestern university. There were two conditions, a positively framed condition and a
negatively framed condition. A total of 122 students participated in the positive condition and
123 students participated in the negative condition. A total of six students were dropped from
the positive condition for failing to complete all required portions of the study. As a result the
final sample consisted of 239 participants, 116 in the positive condition and 123 in the negative
condition. Overall, the sample was 58 percent male and 42 percent female. Participants on
average were 19.6 years of age (sd = 2.1) and had an average ACT score of 25.3 (sd = 3.4). All
students were recruited through the online psychology Human Participation in Research (HPR)
system and received research credit for participation in the study.

Procedure
Students logged into the HPR system using their unique HPR identifier and password.
They then selected the “CRONUS SOLO” experiment from the list of all possible active
psychology experiments. Upon selection of the “CRONUS SOLO” study, students were directed
to an online consent form. After reading the form they electronically indicated their consent (or
opted out of the experiment). At this point, students saw a list of possible in person lab sessions
to sign up for. When students selected a given time and date, they were directed to a series of
online questionnaires. The students then completed surveys for demographics, risk propensity,
goal orientation, and positive and negative affectivity. At the conclusion of the individual
differences measures, students were thanked for their participation and reminded of the in person
42

lab time and date that they signed up for. The entire online portion of the experiment took
approximately 30 minutes for which the students received one research credit.
When students arrived for the in-person lab session they were greeted by the
experimenter, signed in, and were assigned a lab computer to work on. Students were randomly
assigned to either the positive or negative frame condition. When all students arrived, they
watched a short, approximately fifteen minute, training video on the basic functions as well as
the rules and objectives of the computer task, CRONUS SOLO. At the conclusion of the training
�
number and completed up to 35 (𝑇 ≈ 31), five minute scenarios. After completion of all

video, the participants logged into the computer task with their unique university identification

scenarios (or participation for 2.5 hours), the students were thanked for their participation,
debriefed as to the purpose of the present research project, and assigned the remainder of their
research credits (5).

CRONUS SOLO
Crisis Relief Operation Naval Unit Simulation (CRONUS) was originally designed to test
a process model that was derived from the Fiore et al. model of Team Macrocognition (Fiore,
Rosen, Smith-Jentsch, & Salas, 2010). CRONUS simulates a crisis relief effort that the Navy
and other government organizations engage in after a natural disaster such as the devastating
earthquake in Japan in 2011. It was built for a three person team with distributed expertise. As
such, information was divided among three roles, Transport, Intel, and Engineer. For the
purpose of this study, the task was amended for a single player, called CRONUS SOLO, where
all information was available to each individual. For each scenario, individuals were presented
with a map that had a number of potential routes from a starting point to some ending

43

destination. Individuals had to search the map to discover potential obstacles on each route.
Each obstacle had a given value for how much of an effect it had on the total time required to
traverse the route. After determining the scores for each route, the individual must select a route
to have his/her squad traverse the given terrain. The goal for each scenario was to choose the
most efficient route (i.e., lowest score) in the least amount of time possible. There were four
different game screens in CRONUS SOLO, each with information relevant to selecting the best
route. The screens were the Specialist Map, Mission Info, Specialist Info, and Mission
Command. The very bottom of each game screen had tabs with the names of all four screens to
allow the participants to navigate between the screens. This was analogous to navigating
multiple spreadsheets within a Microsoft Excel document.
The Specialist Map screen, shown in Figure 4, displayed an image of a map with a
specified number of routes drawn on it. The map was divided into a 3 by 3 grid. Additionally,
on the right side of the screen there was an information center on the top part of the screen and a
bank of all possible obstacles on the bottom part of the screen. Clicking in a grid square on the
map revealed the bank of potential obstacles and a text link in the information center located on
the top right corner of the game screen that notes if any obstacles were present on the routes
passing through that grid square. Each obstacle had its own unique text link, an example of
which would be “Obstacle on route A.” If obstacles were present in the grid square and the text
link appears, individuals could click on the link to reveal more information about the obstacle.
Left clicking and holding down the mouse button revealed where on the route the obstacle was
located (via an obstacle icon displayed on the map) as well as what obstacle was there (via text
description and unique obstacle icon). An example of the text revealed by clicking on the

44

obstacle link is “Route passes through a water obstacle” coinciding with a picture of the obstacle.
Releasing the left mouse button closed the link and hid all obstacle information.
Participants were informed that to make an obstacle permanently visible on the map, so
that they could easily remember what was there, they may post that obstacle to the route where
the obstacle was located. Choosing to make an obstacle permanently visible equated to the
individual attending to that piece of information because it was an indication that the obstacle
would be incorporated into the final decision. Individuals could post obstacles by clicking on the
relevant obstacle icon in the icon bank at the bottom right corner of the game screen. Clicking
on an icon revealed the option to post that obstacle to any route passing through the selected grid
square. Choosing the appropriate route made the obstacle icon permanently visible on the map
in the location specified by clicking on the text link. Making obstacles permanently visible
allowed participants to easily determine what obstacles are on each route when they had to
calculate total route scores and decide which route is most efficient.
The goal on the Specialist Map screen was to locate all relevant obstacles by clicking in
each grid square on the map, determining what obstacles were on each route in that grid square,
and then posting those obstacles to the map so that they became permanently visible. Performing
these steps allowed participants the best chance to make efficient, effective decisions.

45

Figure 4. Specialist Map screen for CRONUS SOLO. For interpretation of the references to
color in this and all other figures, the reader is referred to the electronic version of this
dissertation.
The next game screen was the Mission Info screen, seen in Figure 5. This screen served
as a training manual and displayed a list of text links with the names of each Microsoft
PowerPoint slide from the training participants took part in prior to engaging in the task.
Clicking on a text link displayed a picture of the PowerPoint slide. Participants could then close
the picture by clicking the “Close” button or by choosing a different game screen by selecting a
different tab at the bottom of the screen. This screen was designed to supplement the initial
training and be a reference for participants if they forgot how to engage in any of the basic
functions of the task.

46

Figure 5. Mission Info screen for CRONUS SOLO.
As seen in Figure 6, the Specialist Info screen displayed information about the point
value of each obstacle. The screen had a list of text links with the names of each obstacle. Left
clicking and holding the mouse button on a link revealed the obstacle icon as well as the point
value for that obstacle (higher numbers indicated more costly obstacles). Releasing the left
mouse button hid the obstacle text. This screen allowed participants to learn how detrimental
each obstacle was, so that they could accurately calculate predicted route scores and choose the
most efficient route. It is important to note that obstacle point values remained constant across
scenarios so individuals could learn the impact that each obstacle had on the total route time.

47

Figure 6. Specialist Info screen for CRONUS SOLO.
The final game screen in CRONUS SOLO was the Mission Command screen shown in
Figure 7. This screen allowed participants to input predicted route scores for each route as well
as select which route they choose to take. Each route had a designated row in a matrix with the
name of the route. Next to the name of each route was a box where participants input the
predicted route scores based on what obstacles were on each route and the obstacle point values
obtained from the Specialist Info screen. Once participants had entered a predicted score for
each route they selected their route choice by clicking a check box next to the route score of the
preferred route. Once a route had been chosen, they could confirm their route choice by clicking
the “Confirm Personal Choice” button located in the center of the right side of the screen.
Clicking on the confirm route choice button brought up a pop-up box that asked participants

48

whether they were certain that the route selected is the one that they wished to choose.
Confirming their choice by selecting “Yes” ended the scenario.

Figure 7. Mission Command screen for CRONUS SOLO.
At the end of each scenario, participants could choose to look at a feedback page. This
screen provided participants with overall feedback as well as detailed feedback for each obstacle
on each route (shown in Figure 8). At the top of the feedback page was a matrix that displayed
the overall number of obstacles correctly discovered for each route and the correct route choice
compared to the route chosen by the participant. This information provided participants with
general feedback on their overall effectiveness in the prior scenario. Scrolling down the

49

feedback page revealed detailed feedback for each route. For a given route, the score input by
the participant was displayed along with the correct route score based on the obstacles that were
actually present in the game. In addition to route score, every obstacle on each route was
displayed with feedback as to its presence in the game (based on given probability) and whether
the participant correctly posted it (or not) on the specialist map. This information provided
participants with detailed feedback as to what obstacles are commonly in the game and whether
they were correctly identifying and posting those obstacles.

Figure 8. Feedback screen for CRONUS SOLO.

50

Scenario Design
Each of the 35 scenarios that participants engaged in was designed with the same basic
architecture. A unique map displayed three routes that traversed the terrain from a starting point
to a final destination. In some scenarios the destination point was the same for all routes; in
other scenarios the destination was different for one or all routes. Prior to the location of any
obstacles, all routes were considered to be equally efficient. The number of obstacles on each
route changed across scenarios; however, the number of overall relevant obstacles within a given
scenario remained constant at 18. Some scenarios also included distracter obstacles in addition
to the 18 relevant ones. These obstacles were on a route but did not affect the overall efficiency
of the route (i.e., had no point value and were not listed on Specialist Info screen). Within a
given scenario there were three distinct types of obstacles, high probability, moderate
probability, and low probability obstacles. There were six of each type of obstacle within each
scenario; however, they were distributed across routes differently across scenarios. This basic
design allowed each scenario to have the same overall difficulty while at the same time allowing
each scenario to present a unique challenge.
The framing of the obstacle text depended on the condition. In the positively framed
condition, information was presented in a way that made it seem as though the obstacle was not
present and, thus, not affect the overall score (resulting in a more favorable outcome). For
example, the obstacle text read “There is an 80% chance that your squad will NOT encounter
hostile forces in the area.” Alternatively, in the negatively framed condition, the same obstacle
information was phrased in a way that made it seem as though the obstacle was present, thus,
making the route less efficient (resulting in a less favorable outcome). For example, the same

51

obstacle text read “There is a 20% chance that your squad will encounter hostile forces in the
area.”
There were a total of ten different obstacles in the game. Three of those obstacles were
high probability, three were moderate probability, and three were low probability. The
remaining obstacle is the distracter that had a probability of 100 percent and does not affect route
score. The three high probability obstacles had a probability of approximately 0.80 (an 80%
chance of affecting the total route score). The three moderate probability obstacles had a
probability of approximately 0.55 (a 55% chance of affecting the total route score). Finally, the
three low probability obstacles had a probability of approximately 0.20 (a 20% chance of
affecting the total route score). In each scenario, there were six obstacles (except the distracter)
with each given probability.
To test the effect of the impact of an outcome on the pattern of biases in risky decision
making, the impact of each type of obstacle was systematically varied. One of each type of
obstacle had a small impact on overall route preference (i.e., one point). Another of each type of
obstacle had a moderate impact on overall route preference (i.e., four points). Finally, the last of
each type of obstacle had a large impact on overall route preference (i.e., seven points).
Completely crossing impact and probability with both positive and negative framing allowed for
testing each type of prospect about which PT and CPT made predictions. Overall, the nine
different types of relevant obstacles (for both the positive and negative frame condition) are
shown in Table 1.

52

Table 1. Obstacle Properties in CRONUS SOLO
Small Impact

Moderate Impact

Large Impact

High Probability (80%)

1

4

7

Moderate Probability (55%)

1

4

7

Low Probability (20%)

1

4

7

Measures
Demographics
Individual differences were assessed using questions on age, sex, academic major, and
year in college.

Cognitive ability
Cognitive ability was assessed through self-report ratings of SAT/ACT scores. To
equalize all participants on a common metric, all SAT scores were converted to the ACT scale.

Affect
To measure positive affectivity and negative affectivity the Positive and Negative Affect
Schedule (PANAS) was administered (Watson, Clark, & Tellegen, 1988). The PANAS uses ten
questions each to assess positive and negative affect. It is a Likert scale with 5 response options
ranging from never (1) to very frequently (5). An example question from the positive affect scale
is “How often to do you generally feel enthusiastic?” Similarly, an example from the negative
affect scale is “how often do you generally feel upset?”

53

Goal Orientation
To measure goal orientation, a 16 item measure of goal orientation developed by
VandeWalle (1997) was administered that utilizes a five point Likert scale from strongly
disagree (1) to strongly agree (5). This scale assessed individuals on the three primary types of
goal orientation: mastery, performance-approach, and performance-avoid. A sample question
from the scale is “I often read materials related to my work to improve my ability.”

Risk Propensity
To assess risk taking propensity, a seven item measure was adopted from the Risk
Propensity Scale developed by Meertens and Lion (2008). Responses were provided on a fivepoint Likert scale that ranges from strongly disagree (1) to strongly agree (5). A sample item is
“I prefer to avoid risks” (reverse coded). The complete scales for all constructs can be found in
Appendix A.

Time Decision Making
To determine whether individuals made decisions more quickly across trials, the actual
time spent making decisions was calculated. The CRONUS SOLO task monitored the total time
that individuals spent left clicking and holding down their mouse button to observe obstacle
texts. The cumulative time spent during each scenario that individuals spent observing this
information was used as a proxy for the amount of time needed to decide whether to post or
ignore an obstacle.

54

Deviations from Optimal Decisions
To determine how far from optimal that individuals made decisions, the overall posting
behavior for each trial was utilized. To calculate optimal decision making, the overall
probability of a prospect (0.80, 0.55, and 0.20) was multiplied by the overall number of each
type of prospect (6). Then, the product of the three obstacle types was added together to
determine the optimal number of posted obstacles per trial (9.3). That number was then divided
by the total number of obstacles per trial (18) to create an overall probability for each scenario
(51.7 percent). If performing optimally, individuals should have posted approximately that
percentage of obstacles. The absolute value of the difference between the percentage of
obstacles posted by an individual and the overall optimal posting percentage was calculated to
determine each individual’s deviation from optimal decision making for each trial.

Risky Decision Making
For each trial, the overall riskiness of decisions was calculated. Similar to the method of
calculating deviations from optimal decisions, the posting behavior of an individual for each
obstacle type (high, moderate, and low probability) within a trial was utilized. Decision making
was considered risky if the percentage posted significantly exceeded the less likely outcome. For
example, decision making was considered risky if an individual posted 35 percent of obstacles
with a probability of 20 percent or posted 65 percent of obstacles with a probability of 80
percent. Similarly, decisions were considered risk averse (or conservative) if the percent posted
significantly exceeded the more likely outcome. For example, decisions were conservative if an
individual posted 90 percent of obstacles with a probability of 80 percent or posted 10 percent of
obstacles with a probability of 20 percent. To calculate a risky decision making score for each

55

person, the percent posted for each type of obstacle was subtracted from the more likely outcome
of the obstacle probability. Then, the three obstacle types were summed to create an overall
risky decision making score for each trial.

Performance
Individual performance for each trial was calculated by determining the number of
obstacles that the individual correctly identified. More specifically, that is the number of
obstacles that the individual posted that were actually present combined with the number of
obstacles that the individual ignored that were actually absent. Means, standard deviations, and
inter correlations among the focal variables can be found in Table 2.

56

Table 2. Correlations among Focal Variables
1. Gender
2. Age
3. ACT

Mean
0.58
19.55
3.06

SD
2.10
0.56

1
-.046
.062

4. Learning GO

3.22

0.61

.058

5. Performance
GO
6. Avoidance
GO
7. Negative
Affect
8. Positive
Affect
9. Riskiness

3.06

0.62

2.83

10. Risky
Decision
#
Making
11. Performance
ŧ

3

-.195*

-

.087

.159*
-.040

.114
-.006

.122

0.72

-.049

-.098

.029

-.363**

.214*

0.80

2.41

0.43

-.085

-.112

ŧ

.079

-.240**

-.027

.162*

0.78

3.40

0.45

.032

-.048

.005

.447**

.121

-.205**

-.419**

2.94

0.64

.141*

-.029

.012

-.101

.001

.002

.015

-.381**
-.016

-0.64

0.16

.164*
-.023

-.001

-0.11

#

2

-.058

.149* 0.80
.055 .046

0.67

0.12

.033

.003

.001

.015

-.025

-.025

-.00

-.013

ŧ

4

ŧ

5

6

7

8

9

10

11

0.85

#

ŧ

0.70

ŧ

0.80

.00

-

-.443**

Note. p < 0.10; * p < 0.05; ** p < 0.01, time-varying covariates were averaged across time points, optimal decision making and
time decision making were excluded because they were not hypothesized to relate to any other variables, reliabilities of individual
differences variables are displayed on the diagonal.

57

-

Data Analysis
The data generated by the aforementioned design is at two distinct levels of analysis.
Responses to gambles over time are nested within-person while individual differences data are
inherently between-person. To account for the nested nature of the data, Hierarchical Linear
Modeling (HLM) was used (Bryk & Raudenbush, 2002; Singer & Willet, 2003). HLM
distinguishes between predictors at different levels of analysis. For longitudinal data, level 1
predictors are those that vary within a person over time. Level 2 predictors, on the other hand,
are those that vary between people but are constant within a person across time points. As such,
time decision making, total obstacle posts, deviations from optimal decisions, risky decision
making, as well as dynamic performance are level 1 variables. Alternatively, demographics,
cognitive ability, positive and negative affect, goal orientation, and risk propensity are all
constant across time and, thus, are level 2 variables.
To justify the use of HLM, intraclass correlation coefficients (ICC(1)) were calculated for
all level 1 variables. HLM divides the overall variance of the system into within- and betweenperson variance components. The ICC(1) is calculated by dividing the between person variance
by the total variance (within plus between). As seen in Table 3, all time-varying processes
showed a significant amount of between person variability, indicating that HLM is the
appropriate analysis and that there is significant variance to predict.
Table 3. ICC(1) Values of Time-varying Covariates
Time Decision Making
Total Obstacle Posts
Deviance from Optimal
Decisions
Risky Decision Making
Performance

Overall
0.36
0.49
0.32

Positive Condition
0.36
0.44
0.29

Negative Condition
0.31
0.52
0.35

0.58
0.16

0.54
0.13

0.61
0.18

58

While HLM is well adept at determining growth trajectories, it is not well suited to
predict more complex reciprocal relationships over time. It is anticipated that prior performance
likely has a lingering effect on decision making. Conversely, it is also likely that trends in
decision making will affect performance. To estimate this reciprocal relationship and model
potential lag effects between decision making and performance, a Vector Autoregressive (VAR)
model was used because VAR allows for simultaneous predictions of lag relationships over time
(Enders, 1995).

59

RESULTS

Prospect Theory Hypotheses
The first three hypotheses are designed to determine whether the overall trends in risky
decision making during information acquisition are consistent with the predictions of Prospect
Theory and its generalizations. To do so, how people respond relative to the actual (true)
probability is important to understand. Likewise, it is necessary to determine how the two
conditions differ in their decision making trends. As a result, a series of t-tests was administered.
There are two specific types of questions that can be tested with the current design. The first is
the overall pattern across people for each time point and the overall pattern averaged across all
time points. The second question is the proportion of people (averaged across all time points)
that demonstrate the proposed bias.
To answer the first question related to Hypothesis 1a, data were collapsed across people
at each time point in the positively framed condition for all high probability pieces of
information. There are a total of six high probability pieces of information for each person for
each time point with 116 people in the positively framed condition, yielding a sample of
approximately 696 responses for each time point. The proportion of responses that are
considered risk averse was calculated by dividing the total number of responses by the number of
responses that chose the more likely (higher probability) option. That sample value was
compared using a single sample t-test to the true value of 80% to determine if individuals are
more risk averse than the true value would predict.

60

Table 4. Average Proportion of High Probability Information Posted by Trial
Negative Condition
Positive Condition
Comparing Conditions
Trial DF % Posted p value DF % Posted p value DF % Posted p value
230
1
122
0.798
0.945
115
0.733
0.037
-0.065
0.123
188
2
122
0.924
0.000
115
0.886
0.000
-0.038
0.165
184
3
122
0.947
0.000
115
0.921
0.000
-0.026
0.228
232
4
122
0.931
0.000
115
0.940
0.000
0.009
0.655
225
5
122
0.932
0.000
115
0.934
0.000
0.002
0.933
232
6
122
0.916
0.000
115
0.941
0.000
0.025
0.245
228
7
122
0.882
0.000
115
0.892
0.000
0.010
0.690
195
8
122
0.924
0.000
115
0.915
0.000
-0.009
0.665
237
9
122
0.909
0.000
115
0.928
0.000
0.019
0.408
212
10
122
0.925
0.000
115
0.915
0.000
-0.010
0.640
227
11
122
0.925
0.000
115
0.927
0.000
0.001
0.952
229
12
122
0.924
0.000
113
0.933
0.000
0.009
0.652
234
13
122
0.912
0.000
115
0.930
0.000
0.018
0.396
229
14
121
0.929
0.000
114
0.930
0.000
0.001
0.941
232
15
122
0.907
0.000
115
0.908
0.000
0.002
0.944
234
16
122
0.905
0.000
114
0.914
0.000
0.009
0.689
226
17
119
0.899
0.000
113
0.923
0.000
0.024
0.319
231
18
121
0.919
0.000
114
0.925
0.000
0.005
0.815
228
19
121
0.919
0.000
112
0.944
0.000
0.025
0.194
230
20
120
0.902
0.000
112
0.923
0.000
0.021
0.373
208
21
105
0.925
0.000
103
0.941
0.000
0.016
0.476
203
22
105
0.904
0.000
98
0.944
0.000
0.040
0.055
194
23
103
0.913
0.000
96
0.943
0.000
0.030
0.225
187
24
103
0.947
0.000
96
0.928
0.000
-0.019
0.359
187
25
101
0.900
0.000
94
0.928
0.000
0.028
0.283
169
26
98
0.931
0.000
89
0.965
0.000
0.034
0.089
180
27
97
0.895
0.000
83
0.927
0.000
0.032
0.188
173
28
94
0.902
0.000
79
0.915
0.000
0.013
0.660
171
29
92
0.907
0.000
79
0.921
0.000
0.014
0.611
155
30
87
0.920
0.000
72
0.909
0.000
-0.012
0.669
151
31
85
0.924
0.000
70
0.941
0.000
0.017
0.486
149
32
83
0.905
0.000
68
0.925
0.000
0.020
0.451
122
33
83
0.917
0.000
60
0.918
0.000
0.001
0.963
133
34
77
0.921
0.000
56
0.953
0.000
0.032
0.208
102
35
73
0.887
0.000
52
0.918
0.000
0.031
0.319
Note. Comparison values created by Positive condition – Negative Condition

61

Table 4 shows that for all trials except the first, individuals attended to a significantly
greater probability of positively framed information than the true probability would indicate.
Additionally, across all trials individuals attended to positive information with high probability
approximately 92 percent of time, significantly greater than the true probability of 80 percent (p
< 0.001). Therefore, Hypothesis 1a was supported. These results indicate that individuals are
risk averse, or overly conservative, when presented with high probability information framed in a
positive way.
It is also possible to determine the percentage of individuals in the positive condition who
were significantly risk averse. To do so, data were collapsed within each person across trials.
There were six high probability pieces of information for each trial and on average individuals
completed approximately 31 trials, resulting in about 186 decisions for each person. A single
sample t-test was performed on each person comparing his/her distribution of decisions to the
true probability of 80 percent.
Table 5. Proportion of Individuals who engaged in each Category of Posting Behavior for High
Probability Information
Condition
Overall
Positive
Negative

Accurately
Posted
0.130
0.138
0.122

Over
Posted
0.799
0.810
0.789

Under
Posted
0.071
0.052
0.089

Always
Posted
0.063
0.069
0.057

Never
Posted
0.000
0.000
0.000

Table 5 shows the percentage of people who accurately posted the proportion of
information, the percentage that significantly over posted the information (risk averse) and the
percentage of people that significantly under posted the information (risk seeking). Consistent
with Hypothesis 1a, the results indicate that a significant proportion of individuals in the positive
condition (81 percent) were risk averse. Additionally, approximately seven percent of the
62

positive condition were extremely risk averse and always posted the high probability
information, accepting an error rate of approximately 20 percent.
An identical procedure to that described for Hypothesis 1a was used to test Hypothesis
1b. As seen in Table 4, individuals in the negative condition were not significantly different
from those in the positive condition. Contrary to the expectations of Prospect Theory,
individuals did not tend to be risk seeking when presented with high probability information
framed in a negative way. Rather, individuals still tended to be risk averse as demonstrated by
the overall rate of posting which is 91 percent, significantly greater than the true probability of
80 percent (p < 0.001) and not significantly less than the amount posted by individuals in the
positive condition (p > 0.10). However, Table 5 shows that the percentage of individuals who
made risky decisions (under post) was greater in the negative condition (9 percent) than in the
positive condition (5 percent). Similarly, the percentage of individuals who were extremely risk
averse was lower in the negative condition (5.7 percent) than in the positive condition (6.9
percent). Although the trends present in table 5 provide evidence that individuals are more risky
when presented with negatively framed information, Table 6 demonstrates that the observed
relationships are not statistically significant. As a result, despite some trends indicating that
individuals are more risk seeking with negatively framed information, overall, Hypothesis 1b
was not supported.

63

Table 6. Differences in Proportion of High Probability Information Posted between Conditions
Difference
DF
Accurately Posted
0.016
234
Over Posted
0.022
236
Under Posted
-0.038
229
Always Posted
0.012
232
Never Posted
N/A
N/A
Note. Difference values created by Positive condition – Negative Condition

p value
0.715
0.676
0.255
0.703
N/A

Hypotheses 1c and 1d were tested in an identical manner to 1a and 1b using data from
low probability information. Consistent with Hypothesis 1c, when presented with low
probability information that was negatively framed, individuals were risk averse, leading them to
ignore more information than the true probability would suggest. As seen in Table 7, in almost
all trials, individuals in the negative condition posted significantly less than the 20 percent true
probability would suggest. Likewise, the overall percent posted across trials was 14 percent,
significantly less than 20 percent (p < 0.01) and the majority of individuals (68 percent)
exhibited risk averse decision making (see Table 8). As a result, Hypothesis 1c was supported.

64

Table 7. Average Proportion of Low Probability Information Posted by Trial
Negative Condition
Positive Condition
Comparing Conditions
Trial DF % Posted p value DF % Posted p value DF % Posted p value
122
0.104
0.000
115
0.121
0.004
222
0.016
0.633
1
122
0.133
0.001
115
0.131
0.009
220
-0.002
0.951
2
122
0.130
0.001
115
0.098
0.000
232
-0.032
0.297
3
122
0.112
0.000
115
0.099
0.000
224
-0.013
0.666
4
122
0.218
0.402
115
0.165
0.147
233
-0.053
0.101
5
122
0.136
0.003
115
0.088
0.000
236
-0.048
0.099
6
122
0.154
0.025
115
0.105
0.000
236
-0.050
0.072
7
122
0.142
0.011
115
0.080
0.000
233
-0.062
0.036
8
122
0.102
0.000
115
0.069
0.000
223
-0.033
0.135
9
122
0.119
0.000
115
0.096
0.000
236
-0.023
0.322
10
122
0.138
0.002
115
0.091
0.000
233
-0.048
0.059
11
122
0.165
0.128
113
0.104
0.000
231
-0.062
0.038
12
122
0.117
0.000
115
0.065
0.000
216
-0.052
0.025
13
121
0.128
0.001
114
0.074
0.000
229
-0.055
0.041
14
122
0.176
0.249
115
0.109
0.000
234
-0.067
0.014
15
122
0.150
0.015
114
0.109
0.000
235
-0.042
0.127
16
119
0.132
0.001
113
0.086
0.000
224
-0.046
0.083
17
121
0.115
0.000
114
0.072
0.000
212
-0.042
0.105
18
121
0.112
0.000
112
0.052
0.000
200
-0.060
0.006
19
120
0.121
0.000
112
0.086
0.000
231
-0.036
0.155
20
105
0.134
0.002
103
0.082
0.000
203
-0.052
0.062
21
105
0.157
0.077
98
0.086
0.000
188
-0.071
0.016
22
103
0.136
0.003
96
0.065
0.000
177
-0.071
0.005
23
103
0.112
0.000
96
0.060
0.000
176
-0.052
0.040
24
101
0.212
0.577
94
0.130
0.001
195
-0.083
0.008
25
98
0.145
0.022
89
0.065
0.000
172
-0.080
0.006
26
97
0.150
0.039
83
0.085
0.000
172
-0.064
0.034
27
94
0.151
0.066
79
0.088
0.000
172
-0.063
0.070
28
92
0.127
0.004
79
0.075
0.000
169
-0.052
0.101
29
87
0.134
0.008
72
0.068
0.000
152
-0.066
0.028
30
85
0.145
0.047
70
0.082
0.000
149
-0.063
0.063
31
83
0.121
0.001
68
0.099
0.000
149
-0.022
0.490
32
83
0.095
0.000
60
0.060
0.000
142
-0.035
0.204
33
77
0.113
0.000
56
0.067
0.000
130
-0.046
0.176
34
73
0.108
0.000
52
0.069
0.000
116
-0.039
0.165
35
Note. Comparison values created by Positive condition – Negative Condition

65

Table 8. Proportion of Individuals who engaged in each Category of Posting Behavior for Low
Probability Information
Condition
Overall
Positive
Negative

Accurately
Posted
0.121
0.103
0.138

Over
Posted
0.146
0.112
0.179

Under
Posted
0.732
0.784
0.683

Always
Posted
0.000
0.000
0.000

Never
Posted
0.176
0.241
0.114

Contrary to the predictions of prospect theory and Hypothesis 1d, individuals presented
with low probability information framed positively were not risk seeking. Rather, they
consistently demonstrated more extreme risk averse decisions than the negative condition. Table
7 shows that in every trial individuals in the positive condition significantly under posted and
overall they only attended to about 9 percent of the information. Similarly, Table 8 demonstrates
that a reasonably high percentage (24) of individuals were extremely risk averse and never
attended to the information, once again those individuals accepted an error rate of approximately
20 percent. However, as was the case with Hypothesis 1b, the differences between conditions in
the percentage of individuals who were risk seeking, risk averse, and extremely risk averse, were
not significantly different (Table 9).
Table 9. Differences in Proportion of Low Probability Information Posted between Conditions
Difference
DF
-0.048
228
Accurately Posted
-0.035
236
Over Posted
-0.067
233
Under Posted
N/A
N/A
Always Posted
0.102
236
Never Posted
Note. Difference values created by Positive condition – Negative Condition

66

p value
0.014
0.411
0.143
N/A
0.076

Table 10. Average Proportion of Moderate Probability Information Posted by Trial
Negative Condition
Positive Condition
Comparing Conditions
Trial DF % Posted p value DF % Posted p value DF % Posted p value
122
0.619
0.030
115
0.642
0.005
236
0.023
0.609
1
122
0.722
0.000
115
0.703
0.000
236
-0.020
0.622
2
122
0.703
0.000
115
0.744
0.000
236
0.041
0.285
3
122
0.728
0.000
115
0.724
0.000
234
-0.004
0.927
4
122
0.683
0.000
115
0.711
0.000
234
0.028
0.477
5
122
0.691
0.000
115
0.713
0.000
235
0.022
0.589
6
122
0.621
0.006
115
0.671
0.000
235
0.050
0.170
7
122
0.721
0.000
115
0.780
0.000
236
0.059
0.119
8
122
0.680
0.000
115
0.721
0.000
236
0.041
0.310
9
122
0.664
0.000
115
0.724
0.000
236
0.060
0.140
10
122
0.642
0.001
115
0.703
0.000
236
0.060
0.121
11
122
0.646
0.000
113
0.721
0.000
235
0.074
0.045
12
122
0.703
0.000
115
0.733
0.000
236
0.030
0.425
13
121
0.667
0.000
114
0.764
0.000
235
0.097
0.015
14
122
0.653
0.000
115
0.704
0.000
236
0.051
0.149
15
122
0.614
0.035
114
0.687
0.000
236
0.073
0.084
16
119
0.606
0.059
113
0.702
0.000
232
0.096
0.019
17
121
0.661
0.000
114
0.739
0.000
235
0.078
0.050
18
121
0.675
0.000
112
0.727
0.000
232
0.052
0.206
19
120
0.624
0.018
112
0.684
0.000
232
0.060
0.162
20
105
0.660
0.001
103
0.718
0.000
207
0.058
0.201
21
105
0.645
0.002
98
0.678
0.000
203
0.034
0.427
22
103
0.606
0.096
96
0.694
0.000
199
0.088
0.060
23
103
0.620
0.035
96
0.694
0.000
199
0.074
0.113
24
101
0.598
0.154
94
0.684
0.000
194
0.086
0.073
25
98
0.586
0.294
89
0.698
0.000
185
0.112
0.023
26
97
0.571
0.517
83
0.641
0.006
179
0.069
0.134
27
94
0.660
0.002
79
0.746
0.000
172
0.086
0.072
28
92
0.600
0.169
79
0.694
0.000
169
0.093
0.077
29
87
0.606
0.122
72
0.660
0.006
155
0.054
0.308
30
85
0.576
0.485
70
0.646
0.018
150
0.070
0.194
31
83
0.615
0.067
68
0.662
0.006
144
0.047
0.375
32
83
0.593
0.256
60
0.642
0.033
133
0.049
0.390
33
77
0.536
0.737
56
0.661
0.020
123
0.124
0.045
34
73
0.514
0.356
52
0.626
0.074
119
0.112
0.052
35
Note. Comparison values created by Positive condition – Negative Condition

67

Although there were not specific predictions about risky behavior for moderate
probability outcomes, it is important to understand the pattern of biases present in decision
making at all levels of probability. Table 10 shows that for moderate probability outcomes,
individuals once again tend to be risk averse and attend to significantly greater amounts of
information than would be suggested by the true probability (55 percent). Despite this general
trend, there were large differences in the pattern of results across the two conditions. As seen in
Table 10, individuals in the negative condition become better over time and began to attend to an
accurate amount of information in later trials. Alternatively, individuals in the positive condition
exhibit no such trend and are overly risk averse the entire time. This difference can also be seen
in the total proportion of information attended to across trials. Individuals in the negative
condition attended to approximately 63 percent of all moderate information while individuals in
the positive condition attend to significantly more (p < 0.05), posting about 69 percent.
Not surprisingly, a similar pattern emerged at the individual level. There were a greater
proportion of individuals in the negative condition that exhibited risky decision making
compared to the positive condition. Conversely, there were significantly more individuals in the
positive condition that were risk averse and extremely risk averse than in the negative condition
(see Table 11). However, these differences were once again not significantly different (Table
12).
Table 11. Proportion of Individuals who engaged in each Category of Posting Behavior for
Moderate Probability Information
Condition
Overall
Positive
Negative

Accurately
Posted
0.213
0.198
0.228

Over
Posted
0.582
0.638
0.528

68

Under
Posted
0.205
0.164
0.244

Always
Posted
0.021
0.026
0.016

Never
Posted
0.000
0.000
0.000

Table 12. Differences in Proportion of Moderate Probability Information Posted between
Conditions
Difference
DF
-0.029
236
Accurately Posted
0.109
236
Over Posted
-0.080
235
Under Posted
0.010
220
Always Posted
Never Posted
N/A
N/A
Note. Difference values created by Positive condition – Negative Condition

p value
0.581
0.087
0.125
0.608
N/A

Hypothesis 2 tested whether the effects of negative framing are stronger than the effects
of positive framing. In cases of both low and high probability information, both conditions
showed similar tendencies to be risk averse. As a result, few significant differences between
conditions were identified. Therefore, Hypothesis 2 was not supported. However, for moderate
information, individuals in the negative condition did show a tendency to attend to significantly
less information, especially with increased experience. Likewise, the percentage of people that
exhibited risky behavior was greater (8 percent more likely) but due to small sample size that
difference was not significant. These results suggest that there are different processes that
individuals engage in due to the framing of information, but those differences may be less
significant at the extremes than Prospect Theory would predict.
Hypothesis 3 examined whether individuals followed the same pattern of risky decision
making for information acquisition as suggested by the weighting function in Cumulative
Prospect Theory. The weighting function suggests that individuals over weight low probability
information, underweight moderate probability information and about accurately rate high
probability information. The decision making observed in information acquisition does not
support this pattern of behavior, thus, not supporting Hypothesis 3. Individuals typically under
weighted low probability information (see Table 7), slightly over weighted and at times
69

accurately weighted moderate probability information (see Table 10), and tended to overweight
high probability information (see Table 4).
A relationship that has received significant attention in the literature is the effect of
outcomes of prior decision making on future decisions. Evidence from theories such as the
house money effect (Thaler & Johnson, 1990) suggest that the riskiness of decisions should
change based on the outcomes of past decisions. More specifically, it is thought that individuals
who experience a prior positive outcome will make more risky decisions while individuals who
experience a prior negative outcome will make more risk averse decisions. It is important to
understand the relationship between prior outcomes and risky decision making because the
riskiness of decisions drastically impacts performance on decision making tasks. Table 13 shows
HLM results from a series of increasingly complex nested models. Models 4 and 5 demonstrate
that when added, decision riskiness significantly predicts performance above and beyond the
effect of experience and the control variables (which were excluded due to non-significance to
conserve degrees of freedom). More importantly, as the riskiness of decisions increases
performance significantly decreases, making it extremely important to understand what can lead
to increased risky decision making.

70

Table 13. Effect of Risky Decision Making on Performance
Model 1
Fixed Effects
Intercept

Model 2

Model 3

Model 4

Model 5

0.666**

0.661**
0.001

0.620**
0.001
0.002
0.007

0.623**
0.000

0.623**
0.000

-0.338**

-0.309**

Trial
Gender
Age
Risky decision making
Random Effects
Residual
Intercept
Time
Risky decision making

0.012
0.002

Note.

0.012
0.002
0.000

0.011
0.000
0.000

0.011
0.002
0.000
0.013

66.29**
3

2

∆Χ
∆df

0.012
0.002
0.000

0.00

947.14**
1

79.47**
3

2

* p < 0.05; ** p < 0.01
Prior research on the relationship between subsequent outcomes on future decisions has

primarily been done using only one or two time points. The current design allows for a more
thorough investigation of how performance and decision riskiness impact one another over time.
To do so, a vector auto-regression (VAR) model was applied to the data. VAR simultaneously
models a reciprocal relationship between prior time points of each variable. More specifically,
both decision riskiness and performance were modeled as a function of their own prior
realizations as well as prior realizations of the other variable. VAR is applied to only a single
person’s data, so all results presented are the average relationship across the entire sample.
When decision riskiness was regressed onto its prior time point and the prior time point
of performance, a few common relationships emerged. Overall, the model explained a
2

significant amount of variance in risky decision making. On average the R value was 0.29. In
32 percent of individuals, prior risky decision making predicted subsequent risky decision
71

̅
making. For those 32 percent, the relationship between the two variables was positive ( 𝛽1 =

0.22), indicating that individuals who made more risky decisions on previous trials made more
risky decisions on subsequent trials as well. In only 7 percent of individuals was the relationship

̅
performance was negatively related to risky decision making ( 𝛽2 = -0.07), suggesting that after
between prior performance and subsequent decision making significant. In those 7 percent,

prior poor performance individuals take a more risky strategy. While it is also possible that after
receiving a positive outcome, individuals take a less risky strategy, it is unlikely because in
general individuals tend to be so risk averse that becoming less risky would be extremely
difficult due to a ceiling effect. These results do not support Hypothesis 4, but are partially
consistent with the escalation of commitment literature (Staw, 1976, 1981) that proposes that
after prior negative outcomes, individuals become more risky in an attempt to perform better.
Performance was also regressed on its prior realizations as well as prior realizations of
risky decision making. Overall, the model did not fit the data particularly well, only explaining
an average of about 9 percent of the variance in performance. Prior time points of performance
only significantly predicted future time points for 5 percent of individuals, about what would be
expected by chance. Prior risky decision making predicted subsequent performance for about 8
percent of individuals. It is likely that prior risky decisions are significantly related to
performance because prior risky decisions are significantly related to subsequent risky decisions
as demonstrated above and risky decisions are related to performance at the current time point
(Table 13).
Although Hypothesis 4 was not supported, an important relationship was discovered.
The pattern of results across the VAR model and HLM model suggests that for some individuals
a reciprocal relationship between risky decision making and performance exists. For those
72

individuals, performance at time t-1 impacts risky decision making at time t, which subsequently
impacts performance at time t.

Individual Differences Hypotheses
Many individual difference variables have been linked to the decision making process.
One of the most commonly studied relationships is between affect and riskiness of decisions.
Hypothesis 5a tested whether there was a main effect of negative affect on risky decision making
such that those higher on negative affect would make riskier decisions. Table 14 presents results
from an HLM model where negative affect was inserted as a predictor. Models 1 through 3 are
testing the unconditional means model, unconditional growth model, and control variables.
Model 4 tests whether the insertion of negative affect predicts differences in risky decision
making. The fixed effect estimate of negative affect (-0.027) was marginally significant (p <
0.10), indicating that on average, individuals higher in negative affect make significantly less
risky decisions, contrary to the expectations presented above. Therefore, Hypothesis 5a was not
supported.
Hypothesis 5b tested whether individuals higher in negative affect would perform worse
overall. As demonstrated in Table 15 (Model 4), negative affect did not significantly predict
differences in performance. As a result, Hypothesis 5b was not supported.

73

Table 14. Effect of Negative Affectivity on Risky Decision Making
Model 1
Fixed Effects
Intercept

Model 2

Model 3

Model 4

-0.110**

-0.112

-0.016

0.002

0.000

0.000
ŧ
-0.028
-0.004

0.000
-0.031*
-0.005
ŧ
-0.027

0.009
0.015
0.000

0.009
0.015
0.000

0.009
0.014
0.000

795.88**
3

0.00

2.34

2

1

Trial
Gender
Age
Negative Affectivity
Random Effects
Residual
Intercept
Time

0.010
0.015

2

∆Χ
∆df

ŧ

Note. p < 0.10, * p < 0.05; ** p < 0.01

Table 15. Effect of Negative Affectivity on Performance
Model 1
Fixed Effects
Intercept

Model 2

Model 3

Model 4

0.666**

0.661**
0.001

0.620**
0.001
0.002
0.007

0.623**
0.000

Trial
Gender
Age
Negative Affectivity
Random Effects
Residual
Intercept
Time
2

∆Χ
∆df

0.001

0.012
0.002

0.012
0.002
0.000

0.012
0.002
0.000

0.011
0.000
0.000

66.29**
3

0.00

0.02

2

1

Note. * p < 0.05; ** p < 0.01

74

It is also important to understand the effect that positive affect has on risky decision
making. Hypothesis 5c investigates whether individuals higher in positive affect attend to more
positively framed information. To test Hypothesis 5c a series of increasingly complex HLM
models was applied to only data from the positive condition (N = 116). Table 16 (Model 4)
shows that the relationship between positive affectivity and information attended to was not in
the hypothesized direction (positive) and the effect failed to be significant. Therefore,
Hypothesis 5c was not supported.
Table 16. Effect of Positive Affectivity on Positive Information Attended
Model 1

Model 3

Model 4

10.302**

10.310**
0.001

10.890**
0.001
0.292
-0.038

11.032**
0.001
0.31
-0.046
-0.292

3.323
2.618

2.709
4.350
0.008

2.718
4.380
0.008

2.717
4.316
0.008

459.79**
3

Fixed Effects
Intercept

Model 2

130.8**
2

0.677

Trial
Gender
Age
Positive Affectivity
Random Effects
Residual
Intercept
Time
2

∆Χ
∆df

ŧ

1

Note. p < 0.10, * p < 0.05; ** p < 0.01

The final proposed relationship between affect and risky decision making is that
individuals higher in positive affectivity will be lower in risky decision making (Hypothesis 5d).
The relationship between positive affect and risky decision making is once again not in the
proposed direction (negative) and is not significant (Table 17, Model 4). So, on average

75

individuals higher in positive affect tended to exhibit riskier behavior, the effect was not
statistically significant. As a result, Hypothesis 5d was not supported.
Table 17. Effect of Positive Affectivity on Risky Decision Making
Model 1

Model 3

Model 4

-0.110**

-0.112

-0.016

-0.020

0.000

0.000
ŧ
-0.028
-0.004

0.000
ŧ
-0.029
-0.004
0.024

0.009
0.015
0.000

0.009
0.015
0.000

0.009
0.014
0.000

795.88**
3

Fixed Effects
Intercept

Model 2

0.00

2.06

2

1

Trial
Gender
Age
Positive Affectivity
Random Effects
Residual
Intercept
Time

0.010
0.015

2

∆Χ
∆df

ŧ

Note. p < 0.10, * p < 0.05; ** p < 0.01

Research has found that when individuals’ goal orientation is manipulated, patterns of
risky decision making change in systematic ways. To date, it is unknown whether the same
systematic relationships exist between trait goal orientation and risky decision making.
Hypothesis 6a examines whether individuals higher in performance-avoid goal orientation make
significantly less risky decisions. Table 18 shows that although the proposed relationship is in
the correct direction (negative), it was not significant. Therefore, while individuals did tend to
slightly be less risky when higher in performance avoid goal orientation, the effect was minimal.
As a result, Hypothesis 6a was not supported.

76

Table 18. Effect of Performance Avoid Goal Orientation on Risky Decision Making

Model 1

Model 3

Model 4

-0.110**

-0.112

-0.016

-0.012

0.000

Fixed Effects
Intercept

Model 2

0.000
ŧ
-0.028
-0.004

0.000
ŧ
-0.029
-0.004
-0.006

0.009
0.015
0.000

0.009
0.015
0.000

0.009
0.015
0.000

795.88**
3

0.00

0.31

2

1

Trial
Gender
Age
Perf. Avoid GO
Random Effects
Residual
Intercept
Time
2

0.010
0.015

∆Χ
∆df
ŧ
Note. p < 0.10, * p < 0.05; ** p < 0.01

Unlike individuals high in performance avoid goal orientation; individuals high in
performance approach goal orientation are thought to exhibit riskier behavior, making them more
likely to engage in risky decision making. Table 19 (Model 4) shows that the relationship
between performance approach goal orientation and risky decision making was trivial and nonsignificant. Therefore, Hypothesis 6b was not supported.

77

Table 19. Effect of Performance Approach Goal Orientation on Risky Decision Making

Model 1

Model 3

Model 4

-0.110**

-0.112

-0.016

-0.016

0.000

0.000
ŧ
-0.028
-0.004

0.000
ŧ
-0.028
-0.004
-0.003

0.009
0.015
0.000

0.009
0.015
0.000

0.009
0.015
0.000

795.88**
3

Fixed Effects
Intercept

Model 2

0.00

0.06

2

1

Trial
Gender
Age
Perf. Approach GO
Random Effects
Residual
Intercept
Time

0.010
0.015

2

∆Χ
∆df

ŧ

Note. p < 0.10, * p < 0.05; ** p < 0.01
A common finding in the decision making literature is that the individual difference trait
of risk propensity is predictive of actual risky decision making. Little evidence of this link exists
from a longitudinal perspective. As a result, Hypothesis 7 examines whether individuals higher
in risk taking propensity make more risky decisions over time. Surprisingly, although the
relationship between the two variables is in the hypothesized direction (positive), the relationship
is not statistically significance (p = 0.12). Therefore, Hypothesis 7 was not supported (see Table
20, Model 4).

78

Table 20. Effect of Risk Taking Propensity on Risky Decision Making
Model 1

Model 3

Model 4

-0.110**

-0.112

-0.016

-0.017

0.000

0.000
ŧ
-0.028
-0.004

0.000
-0.032*
-0.004
0.018

0.009
0.015
0.000

0.009
0.015
0.000

0.009
0.014
0.000

795.88**
3

Fixed Effects
Intercept

Model 2

0.00

2.24

2

1

Trial
Gender
Age
Risk Taking Propensity
Random Effects
Residual
Intercept
Time

0.010
0.015

2

∆Χ
∆df

ŧ

Note. p < 0.10, * p < 0.05; ** p < 0.01

The final individual difference variable that was measured was cognitive ability. It is
predicted that people higher in cognitive ability will make fewer risky decisions (Hypothesis 8a)
and perform better overall (Hypothesis 8b). HLM models were once again used to determine
whether there is a significant main effect of cognitive ability for either risky decision making
(Table 21) or performance (Table 22). Individuals higher in cognitive ability were found to
make significantly fewer risky decisions (p < 0.10), marginally supporting Hypothesis 8a.
Similarly, those higher in cognitive ability performed significantly better over time (p < 0.05),
supporting Hypothesis 8b.

79

Table 21. Effect of Cognitive Ability on Risky Decision Making

Model 1
Fixed Effects
Intercept

Model 2

Model 3

Model 4

-0.110**

-0.112

-0.016

-0.008

0.000

0.000
ŧ
-0.028
-0.004

0.000
-0.023

0.009
0.015
0.000

0.009
0.015
0.000

0.009
0.015
0.000

795.88**
3

0.00
2

2.86
1

Trial
Gender
Age
Cognitive Ability
Random Effects
Residual
Intercept
Time

0.010
0.015

2

∆Χ
∆df

ŧ

Note. p < 0.10, * p < 0.05;

-0.005
ŧ
-0.005

ŧ

** p < 0.01

Table 22. Effect of Cognitive Ability on Performance
Model 1
Fixed Effects
Intercept

Model 2

Model 3

Model 4

0.666**

0.661**
0.001

0.620**
0.001
0.002
0.007

0.615**
0.000

Trial
Gender
Age
Cognitive Ability
Random Effects
Residual
Intercept
Time

0.002*

0.012
0.002

2

∆Χ
∆df

0.012
0.002
0.000

0.012
0.002
0.000

0.011
0.000
0.000

66.29**
3

0.00

3.11*
1

Note. * p < 0.05; ** p < 0.01

80

2

Learning and Decision Making Hypotheses
The final relationships tested are to determine whether individuals change or improve in
their decision making strategy and process over time. Hypothesis 9 posits that individuals will
perform better over time. This hypothesis is tested in two ways. First, an HLM model was
performed analyzing whether over time, the number of correctly identified obstacles improved
over time (Hypothesis 9a). Second, a separate HLM model was run that investigated whether
over time individuals posted obstacles at a percentage closer to the actual stated probability.
Table 23 shows results of the unconditional means model and unconditional growth model of
performance over time. Model 2 demonstrates that on average individuals did not significantly
improve in correctly identifying obstacles. However, despite the non-significance of the fixed
effect, the unconditional growth model did significantly fit the data better than the unconditional
means model. This finding indicates that there is significant variability in growth processes,
meaning that individuals have unique growth trajectories. Therefore, Hypothesis 9a is partially
supported because there is evidence of significant trends in performance over time, but it is not
the consistent improvement that was hypothesized.
A post hoc analysis was conducted using cognitive ability as a predictor to determine
whether there was either a main effect or moderation of ability on performance over time.
Models 3 and 4 in Table 23 demonstrate that while there was, not surprisingly, a marginally
significant main effect of ability, there was no moderating effect. These results indicate that
while overall individuals higher in cognitive ability perform better, their performance trajectories
are not significantly different.
Table 24 shows the unconditional means model and unconditional growth model for
deviations from true probability. Model 2 highlights that over time, individuals became better at

81

posting close to the true probability (p < 0.05), supporting Hypothesis 9b. Taken together,
Hypothesis 9a and 9b demonstrate that over time the decision making process changes and
individuals do become better decision makers, providing evidence of learning. These hypotheses
also suggest that some individuals are more effective learners than others. Once again a post
hoc analysis was performed to determine whether ability significantly predicted differences in
either intercepts or slopes of deviations from optimal decisions. Models 3 and 4 in Table 24
demonstrate that ability does not significantly affect either the initial condition or changes in
optimal decision making.
Table 23. Effect of Time on Performance
Model 1
Fixed Effects
Intercept

Model 2

Model 3

Model 4

0.666**

0.661**
0.001

0.612**
0.000
0.002*

0.604**
0.000
ŧ
0.002
-0.000

0.012
0.002

0.012
0.002
0.000

0.012
0.002
0.000

0.012
0.002
0.000

66.29**
3

0.00

0.20

1

1

Trial
ACT
Trial*ACT
Random Effects
Residual
Intercept
Time
2

∆Χ
∆df

ŧ

Note. p < 0.10, * p < 0.05; ** p < 0.01

82

Table 24. Effect of Time on Deviations from Optimal Probability

Model 1

Model 3

Model 4

0.124**

0.130**

0.144**

0.179**

-0.001*

Fixed Effects
Intercept

Model 2

-0.001
-0.001

-0.003
-0.002
0.000

0.004
0.004
0.000

0.004
0.004
0.000

0.004
0.004
0.000

0.00

2.52

1

1

Trial
ACT
Trial*ACT
Random Effects
Residual
Intercept
Time

0.005
0.002

2

∆Χ
664.51**
3
∆df
ŧ
Note. p < 0.10, * p < 0.05; ** p < 0.01

ŧ

ŧ

The final hypothesis examines whether individuals become more efficient decision
makers over time and potentially automate the process. Table 25 shows the result of an HLM
model designed to test whether the time spent making decision decreases with experience. Not
surprisingly, Hypothesis 10a was supported because over time, individuals on average do
become significantly faster at making decisions (p < 0.01). However, Model 2 also demonstrates
that there is a significant amount of variability in growth trajectories, indicating that there is
significant heterogeneity in the dynamic decision making process, indicating that some
individuals become more efficient over time than others.

83

Table 25. Effect of Time on Time to Make Decisions
Model 1

Model 3

Model4

30.308**

46.885**

44.977**

50.708**

-1.071**

-1.08**
0.086

-1.355**
-0.140
0.011

125.416
278.140
0.356

125.262
281.745
0.340

125.259
281.219
0.339

4220.70**
3

Fixed Effects
Intercept

Model 2

2126.30**
1

0.74

Trial
ACT
Trial*ACT
Random Effects
Residual
Intercept
Time

241.730
136.300

2

∆Χ
∆df

Note. * p < 0.05;

1

** p < 0.01

The final hypothesis (10b) posited that individuals would begin to automate the decision
making process over time, thus beginning to make the same decision for all similar prospects.
This hypothesis was not supported. As evidenced through Hypothesis 9b, individuals are
continuing to improve in their decision making throughout the entire experiment, indicating that
they do not settle on one strategy and stick with it. Similarly, individuals in general are not
prone to make identical decisions for each type of prospect (always post or always ignore). It
was very rare (< 10 percent) for individuals to choose to always make the same decision for
identical prospects and even those individuals who did, tended not to do so with moderate
probability information. Therefore, while it is possible that for some types of decisions (e.g., the
BART), individuals will learn and adopt a consistent strategy, it appears that for information
acquisition where each problem appears unique, individuals continually learn and improve. That
said, it is possible that if observed over sufficient time points that individuals may begin to
employ an identical strategy for all identical decisions.
84

DISCUSSION

The information acquisition process is critical to individuals’ success in organizations.
Whether it is for gathering and analyzing information as a United States intelligence officer, for
accurately diagnosing and treating medical conditions as emergency medical personnel, or for
solving any other organizational problem, the decisions made during information acquisition can
drastically affect organizational outcomes. To date, little is known about how individuals make
decisions about what information is most important to acquire. Since it is rare that what
information is most critical to a solution is obvious, the decision to acquire information can be
modeled as a gamble, where there is a given likelihood that each piece is relevant. Prospect
Theory (PT) and its generalizations is the current dominant approach to the study of gambles for
monetary and nonmonetary outcomes.
While the predictions of PT and its generalizations have been well documented for many
types of risky decisions, most involving time and money, they do not perfectly predict the biases
demonstrated by individuals during the information acquisition process. One of the main
contributions of the PT literature above and beyond rational choice models of decision making is
that PT accurately described instances where individuals were not risk averse. During
information acquisition, however, individuals showed a strong tendency to be risk averse under
all conditions, violating many of the PT predictions. These findings were consistent, however,
with some research using PT to model decisions with time as an outcome where individuals tend
to always be risk averse (Leclerc, 1995).
On the other hand, the rational choice models of decision making also do not accurately
predict the biases demonstrated during information acquisition. Rational choice models of

85

decision making (e.g., Atkinson, 1957; Keeney & Raiffa, 1976) do not account for differences in
riskiness based on question framing or probability of relevance. A common finding during the
information acquisition process was that individuals demonstrated slight differences in their
decision riskiness depending on the probability of information relevance and the way in which
the information was framed. More specifically, when information was framed in such as a way
that made it seem as though something bad was going to happen (negative condition),
individuals tended to be less risk averse for high, moderate, and low probability information.
Likewise, a greater percentage of individuals tended to be risk seeking for all information
probabilities in the negative condition. As a result, while the overall tendency was to be risk
averse, those who received negatively framed information were less so and some even became
risk seeking. Interestingly, individuals in the negative condition also tended to be more accurate,
with a greater percentage of individuals accurately posting both moderate and low probability
information. Similarly, for moderate probability information, individuals in the negative
condition became much better over time and began to post at an accurate level in later trials,
unlike those in the positive condition.
Prospect Theory and rational choice models of decision making do not predict that
individuals will accurately weight uncertain information. However, it has been found that some
individuals will accurately weight probabilistic information, thus choosing each prospect
approximately the correct percentage of times (Bruhin, et al., 2010). This finding was replicated
during information acquisition. For all three probability levels, some individuals correctly
attended to the given proportion of information. Likewise, PT and rational choice models say
nothing of improvement over time. However, it has been found that individuals make more
optimal decisions with repeated experience (Aloysius, 2007), a finding that was replicated in

86

some conditions for the information acquisition process. Therefore, it appears that some
individuals, albeit a minority, are well attuned to the true decision weights in risky decision
making and, thus, do not make the commonly found biases. Likewise, it appears that biases
present in decision making may change over time. As a result, these findings demonstrate that
Prospect Theory cannot accurately predict decision behavior in all cases. This is important
because understanding the limitations and boundary conditions of PT and its generalizations
allow for more accurate use of the theory in training manipulations.
PT and its generalizations explain biases in risky decision making by taking into account
the perceptions and psychological representations of gambles that caused people to deviate from
rational choice. It is possible that individuals perceive the gamble of information acquisition in
such a way that differs from reality. However, the psychological representation of the prospects
during information acquisition is not the same as the representation of the prospects during
monetary gambles. This would account for the fact that individuals do not follow the exact
predictions of Prospect Theory but also do not make purely rational decisions. These findings
are important because understanding the common biases during information acquisition and their
potential underlying causes will allow for better training manipulations to be created to make
people more efficient, optimal decision makers.
Decision riskiness during information acquisition was found to significantly affect
performance. As a result, it is important to understand the conditions under which individuals
are more or less risky so that training programs can be designed to help individuals become more
efficient, optimal decision makers. Despite the fact that Prospect Theory did not completely
describe the biases in risky decision during information acquisition, it did provide a solid
foundation to build upon for discovering the factors that influence risk taking. For example, the

87

likelihood of relevance of a piece of information affected the risky decision making process.
Likewise, the framing of information altered the process by which individuals made risky
decisions. While discovering these two factors and how they impact decision riskiness is an
important first step, additional research is needed to determine other factors that can affect the
sequential risky decision making process during information acquisition.
The relationship between risky decision making and performance is likely not
unidirectional. Theories such as the house money effect (Thaler & Johnson, 1990) and
escalation of commitment (Staw, 1976) highlight the potential importance of prior decision
outcomes (performance) on subsequent risky decisions. Although prior research has supported
both the house money effect and escalation of commitment, it was thought that after performing
poorly individuals would become more risk averse and after performing well, individuals would
become more risk seeking (consistent only with the house money effect). Contrary to these
expectations, it was found that for some people, after performing poorly, they began to make
riskier decisions in an effort to increase performance (consistent with escalation of commitment).
As a result, the timing and presentation of feedback are likely important components of the risky
decision making process. Further research is needed to determine the effects of feedback and the
manner in which feedback is presented on the riskiness of subsequent decisions.
There are likely many potential explanations for why the relationship between prior
performance and decision riskiness was not in the expected direction. Most notably, almost all
prior research on both the house money effect and escalation of commitment has been done
looking at only two or three decisions. The current design used Vector Autoregressive (VAR)
modeling to determine whether across all trials (mean = 31), the relationship between prior
performance and subsequent risky decision making was significant for a given individual. This
88

is a much more thorough test of the proposed relationship than in past studies and it is possible
that while after one or two trials individuals followed the predictions of the house money effect,
overall they altered their behavior when performing poorly by taking more risks in an attempt to
perform better.
Many prior studies on risky decision making have utilized manipulations to determine
whether affect and goal orientation had an effect on static decision riskiness. In organizations, it
is less likely that manipulations for either of these constructs would be very useful during the
information acquisition process. For workers such as the U.S. intelligence officer and
emergency medical personnel, their entire job consists of acquiring information and making
critical decisions under risk and uncertainty. Under these conditions, it would be nearly
impossible to scientifically manipulate employee affect or goals on a consistent basis. As a
result, it is important to determine how the trait forms of affect and goal orientation affect risky
decision making during information acquisition over time.
Despite rather consistent results for the effect of affect and goal orientation on static risky
decisions when manipulations were utilized, the traits of affect and goal orientation did not have
a strong effect on risky decision making over time. Those higher in negative affectivity and
performance avoid goal orientation did show a slight tendency to be overall more risk averse.
Meanwhile, positive affectivity and performance approach goal orientation had no effect on
either performance or decision riskiness. While these findings did contradict expectations, they
are not without precedent. It has been shown that for organizational decisions, individual
difference characteristics have minimal effects on decision making (e.g., Cooper & Wood,
1974). Similarly, affect and goal manipulations are likely to result in much more potent effects,
especially in the short term, than personality traits would demonstrate on long run decision
89

making patterns. These results are once again consistent with the idea that the short term and
long term decision making biases and strategies used by individuals are not necessarily the same
(e.g., Aloysius, 2007) and, thus, the findings and predictions from the Prospect Theory literature
may not accurately reflect the dynamic decision making process. Despite affect and goal
orientation not significantly predicting variability in risky decision making, there is significant
variance left to be explained. As a result, further research is needed to determine what individual
difference characteristics result in significantly different risky decision making behavior.
There is obvious conceptual overlap between the trait risk taking propensity and risky
decision making. Not surprisingly, evidence has shown that those higher in risk taking
propensity make more risky decisions (e.g., Taylor & Dunnette, 1974). However, the
relationship between risk taking propensity and risky decision making over time was not
significant despite being in the proposed direction. While this finding is counterintuitive it is
perhaps not overly surprising. The relationship between risk taking propensity and risky
decision making has been shown to become weakened when decision making is studied over
time (e.g., Sitkin & Weingart, 1995). Similarly, risky decision making behavior has been shown
to differ significantly depending on context (Barseghyan, Prince, & Teitelbaum, 2011), making it
less likely that a stable trait will predict specific behavior. As a result, it is likely that those
higher in risk taking propensity may be more likely to make risky decisions, whether they
actually do will depend strongly on the context and measurement. This means that training
interventions will likely be able to heavily influence the degree to which the predisposition to be
risky actually materializes during information acquisition.
The final individual difference variable that was measured was cognitive ability.
Cognitive ability has been consistently found to be negatively related to risky decision making
90

and positively related to performance. Not surprisingly, these findings were replicated in the
decisions made during information acquisition. As a result, individuals who are higher in
cognitive ability will need less training to become more efficient, optimal decision makers.
As previously mentioned, how the decision making process changes over time has not
received significant attention in the literature. As a result, how people learn and adapt their
decision making strategy is still unknown. Over time, although overall individuals did not
significantly improve in their decision performance, they did begin to attend to information at a
more optimal level. Despite the fact that overall individuals did not significantly increase in their
decision performance, there was significant variability in their slopes over time. This indicates
that some individuals became better over time while others became worse. Further research is
needed to determine what characteristics explain these slope differences. Additionally, the lack
of increase in performance highlights a training or intervention need to help improve decision
accuracy over time. Although there was a significant increase in optimality of decision making
over time, the size of the increase was quite small. Therefore, further research is needed to
determine training interventions to help boost optimal decision making at a quicker rate.
Additionally, the time that individuals take during decision making did significantly decrease
over time. As a result, individuals do become significantly more efficient decision makers
during information acquisition over time.
Recent work by Pleskac and Wershbale (2011) has looked at whether the decision
making process becomes automated with increased experience. Although decision makers on
the BART task did show a tendency to automate their decisions, this tendency was not present
during information acquisition. While time spent decision making and information attended to
did change over time, the change never leveled off to a consistent pattern as would be expected if
91

individuals automated the decision process. It is possible that on more realistic and complex
decisions such as what information to acquire, people take time and consider each decision
uniquely. Likewise, on decisions during information acquisition, it would be a suboptimal
strategy to always choose the same prospect for all similar decisions, especially for moderate
probability information. Therefore, as individuals become better decision makers they should
take more information into account before choosing a prospect. While individuals can become
faster at doing this, as demonstrated here, it is unlikely that they automate the process.

Limitations and Future Directions
Although the findings presented here are novel and interesting, a number of limitations
exist. The primary limitation is the fidelity of the information acquisition decisions. Presenting
individuals with stated probabilities of the relevance/ reliability of information is highly synthetic
and unlike the real world. The decision was made to state probabilities to be consistent with the
decision paradigm used in the prospect theory literature (e.g., Kahneman & Tversky, 1979;
Tversky & Kahneman, 1992). Future research should increase the fidelity by initially presenting
only qualitative statements about the relevance or reliability of information (see Dusenbury &
Fennema, 1996; Gottleib, Weiss, & Chapman, 2007; Windschitl & Wells, 1996) and then
increase further by only demonstrating the relevance or reliability of information through
feedback (see Budescu, Kuhn, Kramer, & Johnson, 2002; Camilleri & Newell, 2009, Hao,
Pleskac, & Hertwig, 2010; Liu & Colman, 2009; Rakow & Newell, 2010; Rode, Cosmides, Hell,
& Tooby 1999; Trautmann, Vieider, & Wakker, 2011). Both of these scenarios have been
shown to alter the biases in the decision making process. Another limitation related to fidelity is
the use of a laboratory simulation. It is possible that motivation and the decision making process

92

would be different under conditions in an actual organization. Future research should attempt to
replicate these findings using a field sample to determine whether the same biases are present.
Another potential limitation is sample size. Some small effects between the individual
difference variables and dynamic processes (e.g., risk taking propensity and risky decision
making) were rather small and would have likely been found significant with a larger sample.
Similarly, while an average of 31 time points is good in organizational research, many more time
points are needed to apply VAR models with multiple predictors. Since VAR is run on a single
individual’s data, time points become the sample size of interest. It is generally accepted that for
each predictor included in the model, researchers should have approximately 40 time points.
This means that it is likely that the VAR relationships reported were under-powered and
additional models including more continuous predictors could not be run.
The current research is descriptive in nature designed to understand the process that
individuals go through when making decisions during information acquisition and document any
potential systematic biases. Training has been shown to improve individuals’ decision making
performance and reduce the biases present in risky decision making (Barron, Leider, & Stack,
2008). The findings in this paper provide a solid foundation to help researchers design effective
training interventions to improve the decision making process individuals go through during
information acquisition. Further research is needed to determine the effectiveness of training
intentions to improve decision making during information acquisition.
Given the limitations just described, a number of future studies are necessary to fully
understand the phenomenon of risky decision making during information acquisition. Likewise,

93

additional research is needed to better explain the current findings and to apply the relevant
results to create training interventions to improve the information acquisition process.
The natural first step in this process is to improve the fidelity of the risky decisions that
individuals must make when acquiring information. As previously mentioned, it is unlikely that
in the real world employees would be given the reliability of information before making a
decision. Rather, it is significantly more likely that they would either have a general qualitative
understanding of the quality of information or would have no knowledge as to the information
reliability until after the decision was made. To model these changes, two follow up studies will
be conducted. The first will use general qualitative statements as to the source reliability such as
“There is a high likelihood that…” which have been shown to alter the nature and severity of
decision biases. The second study would include no descriptor of source reliability but rather
just have deterministic statements about information and utilize only feedback to inform
participants of the likelihood of information being relevant.
Studying decision types that are more realistic will also help determine whether the
reciprocal relationship between performance and risky decision making is more likely in the real
world than found in this paper. It is possible that because feedback in the current study only
served to confirm the stated probabilities, individuals were less inclined to alter their decision
making process based on it, resulting in only 8% of individuals demonstrating the proposed
reciprocal relationship. When feedback is necessary to uncover the underlying reliability of
information, it is possible that individuals will increasingly use feedback to alter their risky
decision making.

94

Similarly, it is possible to model additional individual difference variables in future
studies in an attempt to explain the observed heterogeneity in the decision making process. It
was discovered that there was a significant amount of variability in the intercepts and slopes of
risky decision making, performance, decision optimality, and efficiency. Although cognitive
ability, affect, goal orientation, and risk taking propensity did not explain this significant
variability, it is likely that there are alternative individual difference variables that could account
for additional between person variance.
Along with studies to improve the fidelity of the information acquisition decisions,
additional research should be done to investigate how changing the source of uncertainty in the
information acquisition process affects decision biases. In the current design, all similar
decisions (i.e., obstacles) had identical probabilities. In reality, it may not be that all similar
decisions have identical probabilities but rather that all decisions based on information from a
specific source have approximately equal probabilities. A follow up study will be conducted
looking at how unreliability in information from the same source (e.g., Alpha) about different
decisions (e.g., obstacles) affects the pattern of decision biases found in the current paper.
Understanding how changing the source of unreliability affects biases is the first step toward
generalizing the current findings to the team level, a common context in organizations.
The hope in discovering the nature and severity of decision biases during information
acquisition is to utilize this information to develop training interventions to improve decision
efficiency, optimality, and performance. After gaining a better understanding of the biases under
different conditions by conducting the aforementioned follow-up studies, additional research will
be conducted to develop effective training manipulations. A series of future studies will be

95

conducted examining the effects of different training, agent-based real time feedback, and
retrospective feedback manipulations on mitigating the biases demonstrated in this paper.

Conclusion
As has been documented with time and monetary decisions, individuals are biased during
the information acquisition process if the information presented is uncertain or unreliable.
Unlike the predictions of Prospect Theory, individuals showed an overwhelming tendency to be
risk averse and changed their decision behavior over time. Despite these overall tendencies,
individuals presented with negatively framed information showed a greater likelihood to be risk
taking and performed better over time under certain conditions. Negative affect, performance
avoid goal orientation, and cognitive ability were found to significantly predict risky decision
making behavior. Alternatively, positive affect, performance approach goal orientation, and risk
taking propensity were not found to significantly impact the decision making process. Overall,
individuals become better decision makers over time, increasing in their efficiency as well as
their optimality in terms of information acquired.

96

APPENDIX

97

APPENDIX A

Individual Differences Scales
Participant demographics and background. As noted below, this includes sex, age, class, selfreported GPA, and information regarding computer experience.
1. Sex: M F
2. Age:_____________
3. Major:______________
4. Class standing:
Freshman Sophomore Junior Senior Other
5. Your GPA:______________; your SAT or ACT score:______________
Negative Affectivity. From Watson, Clark, & Tellegen (1988). Answered on a five-point Likert
scale with Never to Very Frequently anchors, with the question worded, "How often do you
generally feel."
1. Distressed
2. Upset
3. Hostile
4. Irritable
5. Scared
6. Afraid
7. Ashamed
8. Guilty
9. Nervous

98

10. Jittery
Positive Affectivity. From Watson, Clark, & Tellegen (1988). Answered on a five-point Likert
scale with Never to Very Frequently anchors, with the question worded, "How often do you
generally feel."
1. Enthusiastic
2. Interested
3. Determined
4. Excited
5. Inspired
6. Alert
7. Active
8. Strong
9. Proud
10. Attentive
Goal Orientation. From Vandewalle, 1997. Answered on a five-point Likert scale with strongly
disagree to strongly agree anchors.
1. I often read materials related to my work to improve my ability.
2. I am willing to select a challenging work assignment that I can learn a lot from.
3. I often look for opportunities to develop new skills and knowledge.
4. I enjoy challenging and difficult tasks at work where I’ll learn new skills.
5. For me, development of my work ability is important enough to take risks.
6. I prefer work situations that require a high level of ability and talent.
7. I would rather prove my ability on a task that I can do well at than to try a new task.

99

8. I’m concerned with showing that I can perform better than my coworkers.
9. I try to figure out what it takes to prove my ability to others at work.
10. I enjoy it when other at work are aware of how well I am doing.
11. I prefer to work on projects where I can prove my ability to others.
12. I would avoid taking on a new task if there were a chance that I would appear rather
incompetent to others.
13. Avoiding a show of low ability is more important to me than learning a new skill.
14. I’m concerned about taking on a task at work if my performance would reveal that I
had low ability.
15. I prefer to avoid situations at work where I might perform poorly.
16. When I don’t understand something at work, I prefer to avoid asking what might
appear to others to be dumb questions” that I should know the answer to already.
Risk Propensity. Adopted from Meertens & Lion (2008). Answered on a five-point Likert scale
with strongly disagree to strongly agree as anchors.
1. I prefer to avoid risks.
2. I take risks regularly.
3. I really dislike not knowing what is going to happen.
4. Safety first.
5. I do not take risks with my health.
6. I view myself as a risk taker.
7. I usually view risks as a challenge.

100

REFERENCES

101

REFERENCES

Abdellaoui, M. F., Vossman, F., & Weber, M. (2005). Choice-based elicitation and
decomposition of decision weights for gains and losses under uncertainty. Management
Science, 51, 1384–1399.
Abualsamh, R. A., Carlin, B., & McDaniel, R. R. (1990). Problem structuring heuristics in
strategic decision making. Organizational Behavior and Human Decision Processes, 45,
159-174.
Ackert, L. F., Charupat, N., Church, B. K., & Deaves, R. (2006). An experimental examination
of the house money effect in a multi-period setting. Experimental Economics, 9, 5-16.
Akers, H. R., Joyner, C. A., Pezzo, M. V., Nash, J. G., Siegal-Jacobs, K., & Stone, E. (1994).
The psychology of windfall gains. Organizational Behavior and Human Decision
Processes, 59, 331-347.
Alge, B. J., Wiethoff, C., & Klein, H. J. (2003). When does medium matter? Knowledgebuilding experiences and opportunities in decision-making teams. Organizational
Behavior and Human Decision Processes, 91, 26-37.
Allias, M. (1953). Le comportement de l’homme rationnel devant le risqué, critique des
postulates et axioms de l’ecole Americaine. Economtetrica, 21, 503-546.
Aloysius, J. A. (2007). Decision making in the short and long run: Repeated gambles and
rationality. British Journal of Mathematical and Statistical Psychology, 60, 61-69.
Andrade, E. B., & Ariely, D. (2009). The enduring impact of transient emotions on decision
making. Organizational Behavior and Human Decision Processes, 109, 1-8.
Ansic, D., & Keasey, K. (1994). Repeated decisions and attitudes to risk. Economics Letters, 45,
185-189.
Ariely, D., & Zakay, D. (2001). A timely account of the role of duration in decision making.
Acta Psychologica, 108, 187-207.
Arkes, H. R. & Blumer, C. (1985). The psychology of sunk costs. Organizational Behavior and
Human Decision Processes, 35, 124-140.
Aronson, E. (1968). Dissonance theory: Progress and problems. In R. Abelson, E. Aronson, W.
McGuire, T. Newcomb, M. Rosenberg, & P. Tannenbaum (Eds.), Theories of cognitive
consistency. Chicago, Illinois: Rand McNally.

102

Aronson, E. (1972). The social animal. San Francisco, California: Freeman.
Atkins, P. W. B., Wood, R. E., & Rutgers, P. J. (2002). The effects of feedback format on
dynamic decision making. Organizational Behavior and Human Decision Processes, 88,
587-604.
Atkinson, J. W. (1957). Motivational determinants of risktaking behavior. Psychological Review,
64, 359–372.
Bandura, A. (1997). Self-efficacy: The exercise of control. New York, NY: Freeman.
Barnes, J. D., & Reinmuth, J. E. (1976). Comparing imputed and actual utility functions in a
competitive bidding setting. Decision Sciences, 7, 801-812.
Barron, G, Leider, S, & Stack, J. (2008). The effect of safe experience on a warnings’ impact:
Sex drugs, and rock-n-roll. Organizational Behavior and Human Decision Processes,
106, 125-142.
Barseghyan, L., Prince, J., & Teitelbaum, J. C. (2011). Are risk preferences stable across
contexts? Evidence from insurance data. American Economic Review, 101, 591-631.
Bereby-Meyer, Y., Meyer, J., & Budescu, D. V. (2003). Decision making under internal
uncertainty: The case of multiple choice tests with different scoring rules. Acta
Psychologica, 112, 207-220.
Bereby-Meyer, Y., Meyer, J., & Flascher, O. M. (2002). Propsect theory analysis of guessing in
multiple choice tests. Journal of Behavioral Decision Making, 15, 313-327.
Betsch, T., Haberstroh, S., Glockner, A., Haar, T., & Fiedler, K. (2001). The effects of routine
strength on adaptation and information search in recurrent decision making.
Organizational Behavior and Human Decision Processes, 84, 23-53.
Blakeney, R. N., & MacNaughton, J. F. (1971). Effects of temporal placement of unfavorable
information on decision making during the selection interview. Journal of Applied
Psychology, 55, 138-142.
Bonner, B. L., Baumann, M. R., & Dalal, R. S. (2002). The effects of member expertise of group
decision-making and performance. Organizational Behavior and Human Decision
Processes, 88, 719-736.
Brehm, J. W., & Cohen, A. E. (1962). Explorations in cognitive dissonance. New York, NY:
Wiley.
Bridgeman, D. L., & Marlowe, D. (1979). Jury decision making: An empirical study based on
actual felony trials. Journal of Applied Psychology, 64, 91-98.

103

Bromiley, P. (2009). A prospect theory model of resource allocation. Decision Analysis, 6, 124138.
Bromiley, P. (2010). Research notes and commentaries looking at prospect theory. Strategic
Management Journal, 31, 1357-1370.
Bruhin, A., Fehr-Duda, H., 7 Epper, T. (2010). Risk and rationality: Uncovering heterogeneity in
probability distortion. Econometrica, 78, 1375-1412.
Budescu, D. V., Kuhn, K. M., Kramer, K. M., & Johnson, T. R. (2002). Modeling certainty
equivalents for imprecise gambles. Organizational Behavior and Human Decision
Processes, 88, 748-768.
Camerer, C. E (1992). Recent tests of generalizations of expected utility theory. In W. Edwards
(Ed.), Utility Theories, Measurement and Applications, Boston, MA: Kluwer Academic
Publishers.
Camerer, C. E & Ho, T. (1994). Violations of the betweenness axiom and nonlinearity in
probabilities. Journal of Risk and Uncertainty, 8, 167–196.
Camilleri, A. R., & Newell, B. R. (2009). The role of representation in experience-based choice.
Judgment and Decision Making, 4, 518-529.
Colquitt, J. A., Hollenbeck, J. R., Ilgen, D. R., LePine, J. A., & Sheppard, L. (2002). Computerassisted communication and team decision-making performance: The moderating effect
of openness to experience. Journal of Applied Psychology, 87, 402-410.
Cooper, M. R., & Wood, M. T. (1974). Effects of member participation and commitment in
group decision making on influence, satisfaction, and decision riskiness. Journal of
Applied Psychology, 59, 127-134.
Cowley, E. (2008). The perils of hedonic editing. Journal of Consumer Research, 35, 71-84.
Cunha, M., & Caldieraro, F. (2009). Sunk-cost effects on purely behavioral investments. Cognitive
Science, 33, 105-113.
De Dreu, C. K. W., & West, M. A. (2001). Minority dissent and team innovation: The importance
of participation in decision making. Journal of Applied Psychology, 86, 1191-1201.
Dean, J. W. & Sharfman, M. P. (1995). Does decision process matter? A study of strategic decisionmaking effectiveness. Academy of Management Journal, 39, 368-396.
Detert, J. R., Trevino, L. K., & Sweitzer, V. L. (2008). Moral disengagement in ethical decision
making: A study of antecedents and outcomes. Journal of Applied Psychology, 93, 374-391.

104

Diehl, E., & Sterman, J. D. (1995). Effects of feedback complexity on dynamic decision making.
Organizational Behavior and Human Decision Processes, 62, 198-215.
Dooley, R. S., & Fryxell, G. E. (1999). Attaining decision quality and commitment from dissent:
The moderating effects of loyalty and competence in strategic decision-making teams.
Academy of Management Journal, 42, 389-402.
Driskell, J. E., & Salas, E. (1991). Group decision making under stress. Journal of Applied
Psychology, 76, 473-478.
Dukerich, J. M., & Nichols, M. L. (1991). Casual information search in managerial decision
making. Organizational Behavior and Human Decision Processes, 50, 106-122.
Dusenbury, R., & Fennema, M. G. (1996). Linguistic-numeric presentation mode effects on risky
option preference. Organizational Behavior and Human Decision Processes, 68, 109-122.
Edwards, W. (1962). Subjective probabilities inferred from decisions. Psychological Review, 69,
109-135.
Enders, W. (1995). Applied econometric time series. New York: John Wiley & Sons.
Etchart-Vincent, N. (2004). Is probability weighting sensitive to the magnitude of consequences?
An experimental investigation on losses, Journal of Risk and Uncertainty, 28, 217–235.
Fellner, W. (1961). Distortion of subjective probabilities as a reaction to uncertainty. Quarterly
Journal of Economics, 75, 670-690.
Festinger L. (1957). A theory of cognitive dissonance. Stanford, CA: Stanford University Press.
Fishburn, P.C., & Kochenberger, G.A. (1979). Two-piece von Neumann-Morgenstern utility
functions. Decision Sciences, 10, 503-518.
Galanter, E., & Pliner, P. (1974). Cross-modality matching of money against other continua. In
H. R. Moskowitz, B. Scharf, & C. Stevens (Eds.), Sensation and measurement. Boston,
MA: Reidel Publishing Company,
Gladstein, D. L., & Reilly, N. P. (1985). Group decision making under threat: The tycoon game.
Academy of Management Journal, 28, 613-627.
Gonzalez, C. (2005). Decision support for real-timer dynamic decision-making tasks.
Organizational Behavior and Human Decision Processes, 96, 142-154.
Gonzalez, R., & Wu, G. (1999). On the shape of the probability weighting function. Cognitive
Psychology, 38, 129–166.

105

Gottlieb, D. A., Weiss, T., & Chapman, G. B. (2007). The format in which uncertainty
information is presented affects decision biases. Psychological Science, 18, 240 -246.
Grayson, C. J. (1960). Decision under uncertainty: Drilling decisions by oil and gas operators.
Cambridge, MA: Graduate School of Business, Harvard University.
Green, P. E. (1963). Risk attitudes and chemical investment decisions. Chemical Engineering, 59.
35-40.
Grove, D. A. (1981). A behavioral consistency approach to decision making in employee selection.
Personnel Psychology, 34, 55-64.
Guzzo, R. A., & Waters, J. A. (1982). The expression of affect and the performance of decisionmaking groups. Journal of Applied Psychology, 67, 67-74.
Halamish, V., Liberman, N., Higgins, E. T., & Idson, L. C. (2008). Regulatory focus effects on
discounting over uncertainty for losses vs. gains. Journal of Economic Psychology, 29, 654666.
Halter, A. N., & Dean, G. W. (1971). Decisions under uncertainty. Cincinnati, OH: South Western
Publishing Co.
Hao, R., Pleskac, T. J., & Hertwig, R. Decisions from experience and statistical probabilities: Why
they trigger different choices than a priori probabilities. Journal of Behavioral Decision
Making, 23, 48-68.
Heath, C., Larrick, R. P., & Wu, G. (1999). Goals as reference points. Cognitive Psychology, 38, 79109.
Helson, H. (1964). Adaptation-level theory. New York, NY: Harper.
Higgins, E. T. (1997). Beyond pleasure and pain. American Psychologist, 52, 1280–1300.
Higgins, E. T. (2000). Making a good decision: Value from fit. American Psychologist, 55,
1217–1230.
Hollenbeck, J. R., Colquitt, J. A., Ilgen, D. R., LePine, J. A., & Hedlund, J. (1998). Accuracy
decomposition and team decision making: Testing theoretical boundary conditions. Journal
of Applied Psychology, 83, 494-500.
Hollenbeck, J. R., Ilgen, D. R., Phillips, J. M., & Hedlund, J. (1994). Decision risk in dynamic twostage contexts: Beyond the status quo. Journal of Applied Psychology, 79, 592-598.
Hollenbeck, J. R., Ilgen, D. R., Sego, D. J., Hedlund, J., Major, D. A., & Phillips, J. (1995).
Multilevel theory of team decision making: Decision performance in teams incorporating
distributed expertise. Journal of Applied Psychology, 80, 292-316.
106

Holloman, C. R., & Hendrick, H. W. (1972). Adequacy of group decisions as a function of
decision-making process. Academy of Management Journal, 15, 175-184.
Horvath, M., Scheu, C. R., & DeShon, R. P. (2001). Goal orientation: Integrating theory and
measurement. Paper presented at the 16th annual conference of the Society for Industrial
and Organizational Psychology, San Diego, CA.
Jackson, S. E. (1983). Participation in decision making as a strategy for reducing job-related
strain. Journal of Applied Psychology, 68, 3-19.
Jackson, S. E. (1984). Correction to “Participation in decision making as a strategy for reducing
job-related strain.” Journal of Applied Psychology, 69, 546-547.
Kahneman, D., & Tversky, A. (1973). On the psychology of prediction. Psychological review,
80, 237-251.
Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk.
Econometrica, 47, 263-292.
Kahneman, D., & Tversky, A. (1984). Choices, values and frames. American Psychologist, 39,
341-350.
Kameda, T., & Davis, J. H. (1990). The function of the reference point in individual and group risk
decision making. Organizational Behavior and Human Decision Processes, 46, 55-76.
Katsikopoulos, K. V., Duse-Anthony, Y., Fisher, D. L., & Duffy, S. A. (2002). Risk attitude
reversals in drivers’ route choice when range of travel time information is provided. Human
Factors, 44, 466-473.
Keeney, R. L., & Raiffa, H. (1976). Decisions with multiple objectives: Preferences and value
tradeoffs. New York, NY: Wiley.
Kleinmuntz, D. N., & Thomas, J. B. (1987). The value of action and inference in dynamic decision
making. Organizational Behavior and Human Decision Processes, 39, 341-364.
Kluger, A. N., Stephan, E., Ganzach, Y., & Hershkovitz, M. (2004). The effect of regulatory focus
on the shape of probability-weighting function: Evidence from a cross-modality matching
method. Organizational Behavior and Human Decision Processes, 95, 20-39.
Kuhberger, A. (1998). The influence of framing on risky decisions: A meta-analysis.
Organizational Behavior and Human Decision Processes, 75, 23-55.
Laughhunn, D. J. & Payne, J. W. (1984). The impact of sunk outcomes on risky choice behavior.
Canadian Journal of Operational research and Information Processing, 22, 151-181.

107

Leclerc, F., Schmitt, B. H., & Dube, L. (1995). Waiting time and decision making: Is time like
money? Journal of Consumer Research, 22, 110-119.
Lee, W. (1971). Decision theory and human behavior. New York, NY: Wiley.
Lejuez, C. W., Aklin, W. M., Jones, H. A., Richards, J. R., Strong, D. R., Kahler, C. W., & Read,
J. P. (2003). The Balloon Analogue Risk Task (BART) differentiates smokers from
nonsmokers. Experimental and Clinical Psychopharmacology, 11, 26–33.
Lejuez, C. W., Aklin, W. M., Zvolensky, M. J., & Pedulla, C. M. (2003). Evaluation of the
Balloon Analogue Risk Task (BART) as a predictor of adolescent real-world risk-taking
behaviors. Journal of Adolescence, 26, 475–479.
Lejuez, C. W., Read, J. P., Kahler, C. W., Richards, J. B., Ramsey, S. E., Stuart, G. L., et al.
(2002). Evaluation of a behavioral measure of risk-taking: The Balloon Analogue Risk
Task (BART). Journal of Experimental Psychology: Applied, 8, 75–84.
Lejuez, C. W., Simmons, B., Aklin, W. M., Daughters, S. B., & Dvir, S. (2004). Risk-taking
propensity and risky sexual behavior of individuals in residential substance use treatment.
Addictive Behaviors, 29, 1643–1647.
LePine, J. A. (2003). Team adaptation and postchange performance: Effects of team composition
in terms of members’ cognitive ability and personality. Journal of Applied Psychology,
88, 27-39.
LePine, J. A. (2005). Adaptation in teams in response to unforeseen change: Effects of goal
difficulty and team composition in terms of cognitive ability and goal orientation.
Journal of Applied Psychology, 90, 1153-1167.
LePine, J. A., Hollenbeck, J. R., Ilgen, D. R., & Hedlund, J. (1997). Effects of individual
difference on the performance of hierarchical decision-making teams: Much more than g.
Journal of Applied Psychology, 82, 803-811.
Liu, H. & Colman, A. M. (2009). Ambiguity aversion in the long run: Repeated decisions under
risk and uncertainty. Journal of Economic Psychology, 20, 277-284.
Locke, E. A. & Latham, G. P. (2002). Building a practically useful theory of goal setting and
task motivation. American Psychologist, 57, 705-717.
MacCrimmon, K. R., & Larsson, S. (1979). Utility theory: Axiomes versus paradoxes. In M.
Allais & G. M. Hagen (Eds), Expected utility hypotheses and the Allais paradox.
Dordrecht, Holland: Reidel Publishing Company.
Markowitz, H. (1952). The utility of wealth. Journal of Political Economy, 60, 151-158.

108

May, K. O. (1954). Intransitivity, utility, and the aggregation of preference patterns.
Econometrica, 22, 1–13.
Meertens, R. M. & R. Lion. (2008). Measuring and individual’s tendency to take risks: The risk
propensity scale. Journal of Applied Social Psychology, 38, 1506-1520.
Michaelsen, L. K., Watson, W. E., & Black, R. H. (1989). A realistic test of individual versus
group consensus decision making. Journal of Applied Psychology, 74, 834-839.
Miner, F. C. (1979). A comparative analysis of three diverse group decision making approaches.
Academy of Management Journal, 22, 81-93.
Nygren, T. E., Isen, A. M., Taylor, P. J., & Dulin, J. (1996). The influence of positive affect on
the decision rule in risk situations: Focus on outcome (and especially avoidance of loss)
rather than probability. Organizational Behavior and Human Decision Processes, 66, 5972.
O’Reilly, C. A. (1977). Supervisors and peers as information sources, group supportiveness and
individual decision-making performance. Journal of Applied Psychology, 62, 632-635.
Ordonez, L., & Benson, L. (1997). Decisions under time pressure: How time constraints affect
risky decision making. Organizational Behavior and Human Decision Processes, 71,
121-140.
Pablo, A. L. (1994). Determinants of acquisition integration level: A decision-making
perspective. Academy of Management Journal, 37, 803-836.
Paese, P. W., Bieser, M., & Tubbs, M. E. (1993). Framing effects and choice shifts in group
decision making. Organizational Behavior and Human Decision Processes, 56, 149-165.
Park, G., & DeShon, R. P. (2010). A multilevel model of minority opinion expression and team
decision-making effectiveness. Journal of Applied Psychology, 95, 824-833.
Permanent Select Committee on Intelligence: House of Representatives. (1996). IC21: The
intelligence community in the 21st century. Retrieved from
http://www.access.gpo.gov/congress/house/intel/ic21/ic21_toc.html
Pietras, C. J., Locey, M. L., & Hackenberg, T. D. (2003). Human risky choice under temporal
constraints: Tests of an energy-budget model. Journal of the Experimental Analysis of
Behavior, 80, 59-75.
Pleskac, T. J. (2008). Decision making and learning while taking sequential risks. Journal of
Experimental Psychology: Learning, Memory, and Cognition, 34, 167-185.
Pleskac, T. J., & Wershbale, A. (2011). Making assessments while taking repeated risk: A
pattern of multiple response pathways. Unpublished manuscript.
109

Quiggin, J. (1982). A theory of anticipated utility. Journal of Economic Behavior and
Organization, 3, 323-343.
Raghunathan, R., & Pham, M. T. (1999). All negative moods are not equal: Motivational
influences of anxiety and sadness on decision making. Organizational Behavior and
Human Decision Processes, 79, 56-77.
Rakow, T., & Newell, B. R. (2010). Degrees of uncertainty: An overview and framework for
future research on experience-based choice. Journal of Behavioral Decision Making, 23,
1-14.
Raudenbush, S. W. & Bryk, A. S. (2002). Hierarchical linear models: Applications and data
analysis methods. Newbury Park, CA: Sage Publications.
Reb, J., & Connolly, T. (2009). Myopic regret avoidance: Feedback avoidance and learning in
repeated decision making. Organizational Behavior and Human Decision Processes, 109,
182-189.
Richter, F. D., & Tjosvold, D. (1980). Effects of student participation in classroom decision
making on attitudes, peer interaction, motivation, and learning. Journal of Applied
Psychology, 65, 74-80.
Ritche, J. B., & Miles, R. E. (1970). An analysis of quantity and quality of participation as
mediating variables in the participative decision making process. Personnel Psychology,
23, 347-359.
Rode, C., Cosmides, L., Hell, W., Tooby, J. (1999). When and why do people avoid unknown
probabilities in decisions under uncertainty? Testing some predictions from optimal
foraging theory. Cognition, 72, 269-304.
Rottenstreich, Y., & Hsee, C. K. (2001). Money, kisses, and electric shocks: On the affective
psychology of risk. Psychological Science, 12, 185-190.
Russo, J. E., & Dosher, B. A. (1983). Strategies for multi-attribute binary choice. Journal of
Experimental Psychology, 9, 676–696.
Schmeidler, D. (1989). Subjective probability and expected utility without additivity.
Econometrica, 57, 571-587.
Scholer, A. A., Zou, X., Fujita, K., Stroessner, S. J., & Higgins, E. T. (2010). When risk seeking
becomes a motivational necessity. Journal of Personality and Social Psychology, 99,
215-231.
Schweiger, D. M., Anderson, C. R., & Locke, E. A. (1985). Complex decision making: A
longitudinal study of process and performance. Organizational Behavior and Human
Decision Processes, 36, 245-272.
110

Seale, D. A., & Rapoport, A. (1997). Sequential decision making with relative ranks: An
experimental investigation of the “Secretary Problem.” Organizational Behavior and
Human Decision Processes, 69, 221-236.
Seo, M., & Barrett, L. F. (2007). Being emotional during decision making – good or bad? An
empirical investigation. Academy of Management Journal, 50, 923-940.
Seo, M., Goldfarb, B., & Barrett, L. F. (2010). Affect and the framing effect within individuals
over time: Risk taking in a dynamic investment simulation. Academy of Management
Journal, 53, 411-431.
Shrode, W. A., & Brown, W. B. (1970). A study of optimality is recurrent decision-making of
lower-level managers. Academy of Management Journal, 13, 389-401.
Singer, J. D., & Willet, J. B. (2003). Applied longitudinal data analysis modeling change and
event occurrence. New York: Oxford University Press.
Singh, J. V. (1986). Performance slack and risk taking in organizational decision making.
Academy of Management Journal, 29, 562-585.
Sitkin, S. B., & Weingart, L. R. (1995). Determinants of risky decision-making behavior: A test
of the mediating role of risk perceptions and propensity. Academy of Management
Journal, 38, 1573-1592.
Slovic, P., & Tversky, A. (1974). Who accepts Savage’s axiom? Behavioral Science, 19, 368373.
Stasser, G. (1992). Information salience and discovery of hidden profiles by decision-making
groups: A “thought experiment.” Organizational Behavior and Human Decision
Processes, 52, 156-181.
Staw, B. M. (1976). Knee-deep in the big muddy: A study of escalating commitment to a chosen
alternative. Organizational Behavior and Human Performance, 16, 27-44.
Staw, B. M. (1981). The escalation of commitment to a course of action. Academy of
Management Review, 6, 577-587.
Steel, P., & Konig, C. J. (2006). Integrating theories of motivation. Academy of Management
Review, 31, 889–913.
Steiner, I. (1972). Group processes and productivity. New York: Academic Press.
Swalm, R. O. (1966). Utility theory – insights into risk taking. Harvard Business Review, 44,
123-136.

111

Taylor, R. N. (1975). Age and experience as determinants of managerial information processing
and decision making performance. Academy of Management Journal, 18, 74-81.
Taylor, R. N., & Dunnette, M. D. (1974). Influence of dogmatism, risk-taking propensity, and
intelligence on decision-making strategies for a sample of industrial managers. Journal of
Applied Psychology, 59, 420-423.
Thaler, R. H. (1980). Toward a positive theory of consumer choice. Journal of Economic
Behavior and Organization,1, 39-60.
Thaler, R. H., & Johnson, E. J. (1990). Gambling with the house money and trying to break
even: The effects of prior outcomes on risky choice. Management Science, 36, 643-660.
Tindale, R. S., Sheffey, S., & Scott, L. A. (1993). Framing and group decision-making: Do
cognitive changes parallel preference changes? Organizational Behavior and Human
Decision Processes, 55, 470-485.
Trautmann, S. T., Vieider, F. M., & Wakker, P. P. (2011). Preference reversals for ambiguity
aversion. Management Science, 57, 1320-1333.
Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases.
Science, 185, 1124-1131.
Tversky, A.& Kahneman, D. (1986). Rational choice and the framing of decisions. Journal of
Business, 59, 251-278.
Tversky, A.& Kahneman, D. (1991). Loss aversion in riskless choice: A reference dependent
model. Quarterly Journal of Economics, 107, 1039-1061.
Tversky, A., & Kahneman, D. (1992). Advances in prospect theory: Cumulative representation
of uncertainty. Journal of Risk and Uncertainty, 5, 297-323
van de Kullen, G., & Wakker, P. P. (2006). Learning in the Allais paradox. Journal of Risk and
Uncertainty, 33, 155-164.
van de Ven, A. H., & Delbecq, A. L. (1974). The effectivenss of nominal, Delphi, and interacting
groups decision making processes. Academy of Management Journal, 17, 605-621.
van Ginkel, W. P., & van Kippenberg, D. (2008). Group information elaboration and group
decision making: The role of shared task representation. Organizational Behavior and
Human Decision Processes, 105, 82-97.
van Ginkel, W. P., & van Kippenberg, D. (2009). Knowledge about the distribution of
information and group decision making: When and why does it work? Organizational
Behavior and Human Decision Processes, 108, 218-229.

112

van Kippenberg, D., van Kippenberg, B., & van Dijk, E. (2000). Who takes the lead in risky
decision making? Effects of group members’ risk preferences and prototypicality.
Organizational Behavior and Human Decision Processes, 83, 213-234.
van Schie, E. C. M., & van der Plight, J. (1995). Influencing risk preference in decision making:
The effects of framing and salience. Organizational Behavior and Human Decision
Processes, 63, 264-275.
VandeWalle, D. (1997). Development and validation of a work domain goal orientation
instrument. Educational & Psychological Measurement, 57, 995-1015.
Viswesvaran, C., & Barrick, M. R. (1992). Decision-making effects on compensation surveys:
Implications for market wages. Journal of Applied Psychology, 77, 588-597.
Vollrath, D. A., Sheppard, B. H., Hinsz, V. B., & Davis, J. H. (1989). Memory performance by
decision-making groups and individuals. Organizational Behavior and Human Decision
Processes, 43, 289-300.
Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief measures
of positive and negative affect: The PANAS scales. Journal of Personality and Social
Psychology, 54, 1063-1070.
Watson, W., Michaelsen, L. K., & Sharp, W. (1991). Member competence, group interaction,
and group decision making: A longitudinal study. Journal of Applied Psychology, 76,
803-809.
Weber, E. U. (1994). From subjective probabilities to decision weights: The effect of asymmetric
loss functions on the evaluation of uncertain outcomes and events. Psychological
Bulletin, 115, 228-242.
Weber, M., & Zuchel, H. (2005). How do prior outcomes affect risk attitude? Comparing
escalation of commitment and the house-money effect. Decision Analysis, 2, 30-43.
Weymark, J. A. (1981). Generalized Gini inequality indices. Mathematical Social Sciences, 1,
409-430.
Williams, A. C. (1966). Attitudes toward speculative risks as an indicator of attitudes toward
pure risks. Journal of Risk and Insurance, 33, 577-586.
Windschitl, P. D., & Wells, G. L. (1996). Measuring psychological uncertainty: Verbal versus
numeric methods. Journal of Experimental Psychology: Applied, 2, 343-364.
Wiseman, D. B., & Levin, I. P. (1996). Comparing risky decision making under conditions of
real and hypothetical consequences. Organizational Behavior and Human Decision
Processes, 66, 241-250.

113

Yaari, M. E. (1987). The dual theory of choice under risk. Econorneotrica, 55, 95-1 15.
Zeelenberg, M., & Beattie, J. (1997). Consequences of regret aversion 2: Additional evidence for
effects of feedback on decision making. Organizational Behavior and Human Decision
Processes, 72, 63-78.
Zeelenberg, M., & van Dijk, E. (1997). A reverse sunk cost effect in risk decision making:
Sometimes we have too much invested to gamble. Journal of Economic Psychology, 18,
677-691.
Zhang, J., Hsee, C. K., & Xiao, Z. (2006). The majority rule in individual decision making.
Organizational Behavior and Human Decision Processes, 99, 102-111.
Ziller, R. C. (1957). Four techniques of group decision making under uncertainty. Journal of
Applied Psychology, 41, 384-388.

114