..ar0:..0!vc.ll.VI..... I
x!!.!!.-.-i1t ill! U-\.

 

 

.46 ’3 :2? ... .v .15 J ‘ 2. .la.....y:pi» rip . ...,... h ,. v4...‘!.fr\.
kaex. ....ru..5u.¢.s..;w_1.w 5.3.2.»... 1:53.... : i» . . - - me: t?

.. ‘ ..c_.,...... 5...‘k....u ....wr.¢.?.ufiush.,nu.¢d.fihflf _.....wu.n..mul,u, v...“
. ‘ ‘ . . ‘ , . . . A . .

 
 

u.. .s ..

 

I ‘ |

 

(r1633

I—
asml U3 \

MI

\\

Win

3

This is to certify that the

dissertation entitled

DECISION MAKING IN INDIVIDUAL AND TEAM CONTEXTS:
MODERATORS OF THE EFFECTS
OF COGNITIVE FRAMES ON RISK TAKING

presented by

DEBRA ANN MAJOR

has been accepted towards fulfillment
of the requirements for

PA . A degree in M1!”

 

)7 cz/Z'a 77%—

ajor profesflfi

Date [Mm/,L

.545“ is an Affirmative Action /Equul Opportunity Institution 042771

 

F LIBRARY
Michigan State

University J

 

 

‘\_

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

—I_:_

MSU Is An Affirmative Action/Equal Opportunity Institution
empire-9.1

 

 

 

 

 

 

 

 

.fi

DECISION MAKING IN INDIVIDUAL AND TEAM CONTEXTS:
MODERATORS OF THE EFFECTS OF COGNITIVE FRAMES ON RISK TAKING
By

Debra Ann Major

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Psychology

1992

ABSTRACT

DECISION MAKING IN INDIVIDUAL AND TEAM CONTEXT S:
MODERATORS OF THE EFFECTS OF COGNITIVE FRAMES ON RISK TAKING

By

Debra Ann Major

An individual model of risky decision making was developed to further
understand the effects of cognitive decision frames on risk taking and the boundaries
of such framing effects. The model took issue with previous research that has shown
that individuals presented with a decision framed as a loss tend to be more risk
seeking than when the same decision is framed as a gain. It was argued that the
differential risk seeking exhibited under gain and loss frames is a phenomenon limited
to the hypothetical, lottery-type, decision tasks commonly employed in research on
the effects of cognitive frames. Perceived control was hypothesized as a key
moderator of the relationship between cognitive decision frames and decision
riskiness. Individuals with lower perceived control were expected to make more risk
averse decisions under gain frames and more risk seeking decisions under loss frames.
Individuals with higher perceived control were predicted to make riskier decisions
regardless of decision frame.

The individual model of risky decision making was used as a basis for the
development of a model of risky decision making in a team context. The team
context model focused specifically on hierarchical teams with distributed expertise and
the effects of cognitive frames on risky decision making in that context. Perceived

control was again hypothesized as a critical boundary condition. In addition, the team

context model described the potential moderating effects of need for affiliation and
team cohesiveness on the relationship between cognitive frames and decision
riskiness.

To test the hypotheses derived from the two models, individuals (I; = 70) and
four-person teams (11 = 41 teams) participated in a computerized decision making
simulation called TIDEZ. Individuals and teams were assigned to either a loss or a
gain condition. Perceived control was operationalized as the perceived control over
the probability of making a correct decision. An attempt was made to create
variance in perceived control by manipulating the ambiguity of information upon
which decisions were based.

Results Showed predicted main effects for framing in both the individual and
team decision making contexts. At the individual level, results supported the
hypothesized moderating effect for perceived control. In the team context, perceived
control had a main effect on decision riskiness, but was not a Significant moderator.
Team cohesiveness, however, was a Significant moderator of the effects of cognitive
decision frames on decision riskiness. The findings were discussed in terms of their

implications for risky decision making in organizations.

In loving memory of special grandparents,

Lewis Maj or and Helen Murphy

iv

ACKNOWLEDGMENTS

A number of people contributed time and talent to make this dissertation
possible. I would like to take this opportunity to thank them. Andy Johanson
developed the basic TIDE2 software and cheerfully worked with me to modify the
Simulation for the present research. Douglas Sego was also instrumental in refining
TIDE2 and establishing the Team Effectiveness Lab at Michigan State University
where this Study was conducted. Doug and I were fortunate to be the first graduate
students assigned to the Team Effectiveness Lab and it has been a pleasure to work
with him from the beginning. A number of other individuals were invaluable in
recruiting subjects and collecting data for this study, including Jean Phillips, Jennifer
Hedlund, Paul Barrett, Linda Barrett, Mark Roehling, and Larry Seese. Their help is
greatly appreciated.

Of course without my co-chairs, Daniel Ilgen and John Hollenbeck, the Team
Effectiveness Lab would not have become a reality. I would like to thank them for
providing the means to do this research and also for the constant encouragement they
provided and the considerable expertise they brought to all aspects of this project.
My committee members, Steve Kozlowski and Kevin Ford, have been friends and

mentors since my undergraduate days at Michigan State. They brought a fresh

perspective to this research. Their insight was extremely valuable in developing the
research and refining the final product.

Finally, I’d like to thank my friends and family for providing moral support
during the ”conception and birth” of this dissertation. Their patience and
understanding throughout the entire graduate school process has been phenomenal. I

hope I can rely on their support throughout my academic career.

vi

TABLE OF CONTENTS

LIST OF TABLES ..................................... xiii
LIST OF FIGURES ..................................... xiv
INTRODUCTION ....................................... l

The Impact of Cognitive Frames on Risky Decision Making . . . . 2
Individual and team context issues ................ 3

THEORIES OF RISKY DECISION MAKING AT THE INDIVIDUAL

LEVEL ......................................... 5
Subjective Expected Utility Theory ................... 6
The axioms of subjective expected utility theory ....... 6
Supportive empirical research .................. 9
Contradictory theoretical and empirical arguments ...... 11
Prospect Theory ............................... 12
The certainty effect ......................... 12

The reflection effect ........................ 14

The reference effect ........................ 17
Decision making according to prospect theory ........ 19
Empirical research ......................... 20
Limitations of empirical research ................ 25

vii

Understanding Rational and Nonrational Models of Decision

Making ................................ 26
Gain frames ............................. 30
Loss frames ............................. 32
Perceived Control .............................. 33

literatures .......................... 33
Perceived control and risk taking ................ 34
Perceived control in the decision making context ....... 35
Perceived control as a moderator ................ 38

Summary: Risky Decision Making, Cognitive Frames, and
Perceived Control .......................... 40

EFFECTS OF COGNITIVE FRAMES AND PERCEIVED CONTROL

ON RISKY DECISION MAKING FOR INDIVIDUALS ........... 42
The Moderating Effect of Perceived Control .............. 42
Factors Influencing Perceived Control .................. 44

Locus of control .......................... 45

Task knowledge ........................... 46

Performance history ........................ 47

Cue ambiguity ............................ 48

Individual Level Hypotheses ........................ 49
viii

 

RISKY DECISION MAKING IN THE TEAM CONTEXT .............. 52

Team Versus Individual Decision Making ............... 52
Teams defined ............................ 52

Hierarchical teams with distributed expertise ......... 54

Empirical Research: Applieation to Hierarchical Teams ........... 5 6
Risky Shift .................................. 56
Individual decisions ........................ 57

Team decisions ........................... 58

Conclusions ............................. 64

Prospect Theory Framing Effects ..................... 65
Leader Decision Making .............................. 70
RISKY DECISION MAKING IN HIERARCHICAL TEAMS ............ 75

Hierarchical Decision Making Teams with Distributed Expertise . . 75

General characteristics ....................... 75

Specific characteristics ....................... 77

Critical Features of the Team Context ...................... 78
Decision Cues ................................ 78
Decision Making Process ......................... 79
Decision Outcomes ............................. 80

The Team Context Model of Risky Decision Making ............. 81
Need for affiliation ......................... 83

Cohesiveness ............................ 84

ix

Predictors of Leader Perceived Control ..................... 91
Applications of the Individual Model .................. 91

Locus of control .......................... 91

Task knowledge ........................... 92

Performance history ........................ 93

Cue ambiguity: Informational and recommendation ..... 94

Predictors of Perceived Control Derived from the Team Context . 95

Centralization ............................ 96

Coordination ............................. 97

Additional Issues Relevant to the Team Context ................ 98

Cue Weighting ................................ 98

Magnitude of Framing Effects ..................... 100

Hypotheses for Risky Decision Making in a Team Context ........ 101
METHOD ........................................... 106
Participants ................................. 106

Design .................................... 106

Decision Task ............................... 106

Mathematical structure underlying the decision task . . . . 108

Feedback .............................. 110
Simulation Configuration and Cue Access .............. 112
Individual version ......................... 112

Team version ........................... 113

Procedure .................................. 117
Procedure for individual participants ............. 118
Procedure for team participants ................ 119

Manipulations ............................... 121
Framing .............................. 121
Ambiguity ............................. 124

Measures .................................. 127
Decision riskiness ......................... 127
Perceived Control ......................... 127
Task knowledge .......................... 128
Locus of control ......................... 128
Performance history ....................... 129
Need for Affiliation ....................... 129
Cohesiveness ........................... 129
Centralization ........................... 129
Coordination ............................ 130

RESULTS ........................................... 131

Test of Individual Hypotheses ...................... 134

Tests of Team Context Hypotheses ................... 143

The Process of Risky Decision Making ................ 150

Comparing Framing Effects in the Individual and Team

Contexts .............................. 153

DISCUSSION ........................................ 155
Summary and Interpretation of Results ................ 155

Individual decision makers ................... 155

Team leaders ........................... 158

Comparison of the individual and team contexts ...... 163

Summary .............................. 164

Contributions and Limitations ...................... 165

Issues of control ......................... 165

Risky decision making in teams ................ 169

Decision accuracy ........................ 170

Concluding Remarks ........................... 171

APPENDDI A: GENERAL INSTRUCTIONS .................... 173
APPENDIX B: PERCEIVED CONTROL ....................... 188
APPENDIX C: TASK KNOWLEDGE TESTS .................... 190
LIST OF REFERENCES ................................. 194

xii

Table 1:
Table 2:

Table 3:

Table 4:

Table 5 :
Table 6:
Table 7:
Table 8:
Table 9:
Table 10:
Table 11:
Table 12:

Table 13:

LIST OF TABLES

Rules that Protect Decision Quality and Acceptance ............
Decision Processes ................................
The Nine Informational Cues and the Regression Equation

for the Correct Decision .........................
Direct Access to Informational Cues and Rules Describing

Interactions among Cues .........................
Decision Outcomes ...............................
Points Awarded under Gain and Loss Frames ...............
Individual Level Correlation Table ......................
Repeated Measures Regression Results for Hypotheses 1 & 2 .....
Repeated Measures Regression Results for Hypothesis 3 ........
Team Context Correlation Table .......................
Repeated Measures Regression Results for Hypotheses 4, 5, & 6 . . .
Repeated Measures Regression Results for Hypothesis 7 ........

Hierarchical Regression Results for Hypothesis 9 ............

xiii

71

114

122

125

135

137

141

144

146

151

Figure 1:
Figure 2:
Figure 3:
Figure 4a:
Figure 4b:
Figure 5:
Figure 6:
Figure 7:
Figure 8:
Figure 9:
Figure 10:
Figure 11:
Figure 12:

Figure 13:

LIST OF FIGURES

A hypothetical value function ......................... 21
Hypothesized moderating effect of control ................. 39
Model of individual risky decision making ................. 43
Illustration of group polarization effects .................. 61
Illustration of group attenuation effects .................. 61
Hypothesized moderating effect of need for affiliation ......... 85
Hypothesized moderating effect of cohesiveness ............. 88
Primary model of team level risky decision making .......... 90
Elaborated model of risky decision making in a team context ..... 102
Performance outcomes ............................ 11 1
Partitioning the between and within subjects variance ......... 132
Moderating effect of perceived control .................. 138
Interaction between task knowledge and locus of control ....... 142
Moderating effect of team cohesiveness .................. 148

xiv

INTRODUCTION

Decision making is a primary topic within the field of industrial/organizational
psychology. A great deal of theoretical and empirical attention has been devoted to
understanding and predicting decision making processes. Within the decision making
domain, risky decision making is a particularly critical topic. Decision alternatives
are rarely certain. That is, the outcomes resulting from a particular decision may be
unknown or only partially known. These are the circumstances that define risky
decision making. Risky decision making can involve either a choice among certain
and probabilistic alternatives or a decision among several alternatives, all of which
have probabilistic outcomes.

From an applied perspective, risky decision making is a Significant area of
research because of its pervasiveness and criticality. Most decisions are probabilistic
in nature. Rather than an exception, risky decision making is the norm. In addition,
most decisions of extreme consequence fall under the rubric of risky decision making.
Doctors, business executives, and politicians are just a few examples of individuals
whose careers are based on making significant risky decisions.

Risky decision making, however, is not solely an individual phenomenon. A
great deal of risky decision making occurs in a team context. In the military, for

instance, command and control teams make crucial decisions regarding defensive

2

actions against hostile forces. Cockpit crews in commercial airliners make decisions
regarding aircraft operation that may have a significant impact on the safety of all
those aboard. In business, management teams make decisions that have critical
implications for firms’ profitability and survival.

Focusing attention on the team context adds an interesting dimension to the
Study of risky decision making. The team context introduces several features (e.g. ,
social processes, hierarchical structures, communication requirements) that may
impinge on risky decision making processes. The unique characteristics of the team
environment are likely to have an impact on risky decision making even if an
individual (i.e., a team leader) is ultimately responsible for the team’s decision. In
other words, the individual risky decision making process iS expected to be
fundamentally altered by virtue of occurring in a team context. The present research
is interested in risky decision making in both individual and team contexts.

i i M

The more specific goal of the present research is to explore the effects of
cognitive frames on risky decision making, in particular those circumstances under
which frames are and are not likely to have a significant impact on decision making
under risk. The notion of "framing” emerged as an issue relevant to understanding
risky decision making in the late 19703 and early 1980s (Kahneman & Tversky, 1979;
Tversky & Kahneman, 1981). The appearance of work related to cognitive frames

coincided with a recognition in the literature that decision behavior often did not

3

conform to the prescriptions of normative theories and perhaps Should not be expected
to meet the assumptions of rational models (Einhorn & Hogarth, 1981).

In the decision making context, a cognitive frame refers to the way in which a
decision problem is couched or presented. Kahneman and Tversky (1979)
incorporated the notion of framing into a model of risky decision making called
prospect theory. The basic proposition of prospect theory is that the way in which a
decision problem or choice is framed can dramatically alter the ultimate decision that
is made. Prospect theory was able to account for observed decision behavior that the
normative rational model (i.e. , expected utility theory) could not.

The present research further explores the impact of cognitive frames on risky
decision making, seeking to extend the literature in at least two respects. First, the
present research examines the effects of cognitive frames in a team context, carefully
considering those aspects of a team most likely to impinge on the process of decision
making under risk. Second, this research attempts to identify some of the boundary
conditions of framing effects by proposing and studying critical moderators of the
relationship between cognitive frames and risky decision making in both the individual
and team contexts.

WM- Most of the theoretical and empirical work
related to risky decision making has been conducted at the individual level. Thus,
that literature is explored first in order to build a conceptual model of risky decision
making at the individual level and to derive hypotheses relevant to individual decision

makers acting alone. The individual risky decision making literature also serves as

4

the foundation for developing a model for leaders engaged in risky decision making in
a team context. Constructs borrowed from the team and leadership literatures are
integrated into a model of risky decision making that serve as the basis for hypotheses

regarding the effects of the team context.

THEORIES OF RISKY DECISION MAKING AT THE INDIVIDUAL LEVEL

This section is devoted to a discussion of two theories of risky decision making
relevant to individuals: subjective expected utility theory and prospect theory. To
understand the impetus for interest in the effects of cognitive frames on risky decision
making, one must first understand the rational model of decision making under risk.
Subjective expected utility theory describes a classical model of decision behavior,
which considers the decision making process from a normative rational perspective.
Dissatisfaction with the adequacy and accuracy of the rational model prompted the
development of prospect theory. The prospect theory model applied the notion of
cognitive decision frames to account for the ”nonrational" violations of subjective
expected utility theory evident in individual decision making.

Although the two theories propose some conflicting hypotheses and each
approaches decision making behavior from a different orientation, the goal is not to
pit one theory against the other. Instead, the position take here is that both theories
are critical to understanding risky decision making. Below, a description and critique
of subjective expected utility theory is provided first, followed by a discussion and
evaluation of prospect theory. AS previously mentioned, it is important to first
comprehend the rational model (i.e. , subjective expected utility theory) before
Studying a model developed to account for departures from it (i.e. , prospect theory).
The section concludes with an attempt to reconcile the two theories by identifying the
ways in which they define each other’s boundaries, especially the boundaries of

cognitive framing effects.

Subjective expected utility theory (SEU) was developed to provide a normative
rational model for individual decision making. Directed mainly at predicting
economic decisions, the primary underlying assumption of the model iS that
individuals’ economic decisions are driven by the desire to maximize utility in terms
of outcomes or profits. Rational behavior is equated with maximizing expected
utility. Individuals’ expected utilities are labelled ”subjective” because SEU is
concerned with individuals’ beliefs or wagons regarding the probabilities of
maximizing outcomes. The mathematical roots of the SEU model are traceable to
Bernoulli (1738). The original theoretical conceptualization is attributed to von
Neumann and Morgenstem (1944, 1947), while Savage (1954) is credited with the
most complete explication of the subjective element.

W. The theory is based on a set
of axioms that describe the expected behaviors of a rational decision maker. The
descriptions of the axioms provided below are adapted from Schoemaker (1980) who
provides a less technical, more applied interpretation of SEU principles (see von
Neumann & Morgenstem, 1947 & Savage, 1954 for mathematical descriptions of the
axioms).

(1) WW: For any two outcomes, a decision maker either
prefers one over the other (i.e., A < B or A > B) or is indifferent (i.e., A =

B). TransitiLitx also exists such that if A is preferred to B and B is preferred

to C, then A will also be preferred to C.

(2)

(3)

(4)

(5)

7
MW: If A is preferred to B and B is preferred to C, then

there must exist some probability (p), ranging from 0 to 1, such that an
individual is indifferent to a choice between a guaranteed B outcome and the
chance to obtain A or C with respective probabilities of (p) and (l-p).
Wm: If a decision maker is indifferent to A and B, the
decision maker will also be indifferent between two lotteries, one offering A
and C with probabilities of (p) and (1-p) and the other offering B and C with
the same probabilities. This is true for any C and (p) value.
MW: If A is preferred to B, a decision maker faced
with a choice between two lotteries which both offer only A and B outcomes
will prefer the lottery which offers the highest probability of obtaining A.
W: If one lottery (Ll) offers outcomes A and B and a
second lottery (L2) has as its outcomes two additional lotteries which both
offer only A and B as outcomes, then a decision maker Should be indifferent
between L1 and L2, if and only if the expected values of L1 and L2 are
identical.

Since SEU theory contends that all individuals adhere to the axioms described

above, the theory also posits that it is possible to use the axioms to derive a utility

function that should predict or describe an individual’s decision behavior for a given

choice problem. In operational terms, subjective expected utility theory describes

individual decision making in the following manner: (a) a given decision has a certain

number of alternatives, (b) certain outcomes are associated with each alternative, (0) a

8

decision maker assesses the value of each outcome (e. g. , the attractiveness or
aversivencss) and the probability that the outcome will be obtained if the associated
alternative is selected, ((1) the decision maker determines the expected worth of each
outcome by multiplying its value by its probability, (e) the expected utility of a
decision alternative is determined by summing the worth of all possible outcomes
associated with it, and (t) the decision maker selects the decision alternative with the
greatest worth, termed subjective expected utility.

Recall that the underlying assumption of the SEU model is that individuals
desire a maximization of outcomes. In some instances, maximizing outcomes means
minimizing losses or aversive events. Therefore, a decision alternative becomes more
attractive as its consequences become more positive or less negative. According to
the theory, highly likely positive outcomes and unlikely negative consequences are
preferable to low probability positive consequences and high probability negative
outcomes.

In terms of risky decision making, SEU does not predict a general tendency to
prefer decision alternatives with certain or probabilistic outcomes. Instead, risky
decisions are proposed to be based on the value an individual assigns to expected
outcomes, combined with the perceived probability that the outcomes will be
forthcoming if the associated decision alternative is selected. While the model does
not deal specifically with an individual’s general attitude toward risk taking, such
attitudes are captured to the extent that individual risk tendencies play a role in

assigning values and interpreting probabilities. For example, even if outcome A is

9

valued over outcome B, an individual’s utility function will predict the selection of
outcome B if the probability associated with it is high enough so that the utility of B
excwds that of A.

For more than fifty years, subjective expected utility theory has been a
principal model of risky decision making. The axioms have an intuitive appeal and
empirical research has supported the efficacy of the utility function in describing and
predicting decision behavior (e. g. , Davidson, Suppes, & Siegel, 1957; Mosteller &
Nogee, 1951; Schoemaker, 1980). The theory, however, has also generated a
considerable amount of controversy based on both theoretical and empirical grounds.
In the sections that follow, classical research supporting the SEU model is described
and the theoretical and empirical arguments refuting the theory are enumerated.

W. Literally volumes could be devoted to
describing the empirical research generated by the subjective expected utility model.
This work can be roughly categorized into studies that, (a) attempted to assess the
viability of SEU theory as a reasonable model of decision making, and (b) those that
sought to refute the theory or individual axioms. Most early empirical research based
on the SEU model of decision making belongs to the former category. This work
focuses on two major goals: (a) ascertaining whether or not utility functions as
described by SEU theory can be derived for individuals, and (b) determining if such
functions can be used to accurately predict individual decision making. Two classical

studies representative of this body of research are described below.

10
Mosteller and Nogee (1951) conducted one of the first experimental tests of

the SEU model. They engaged a sample of college students and National Guard
members in a betting task using small amounts of money. Over repeated trials,
preferences and indifferences for a variety of bets were determined, allowing for the
construction of probabilistic utility functions. Consistent with SEU theory, Mosteller
and Nogee found that bets associated with greater expected utilities had a higher
probability of being selected. Furthermore, utility functions once constructed were
reasonably good predictors of betting decisions and were better predictors of betting
decisions than functions based solely on the monetary value of bets.

Another classic Study conducted by Davidson et a1. (1957) assessed utility
functions for a single chance event. Using indifference judgments, six utility points
were determined for each subject in their student sample. Based on their empirical
results, Davidson et a1. concluded the following: (a) utility functions were generally
not linear, (b) subjects decisions were consistent with maximizing SEU, and (0) based
on remeasurement, subjects appeared to be consistent in their decisions over time.

The two studies described above are illustrative of SEU research that has been
conducted using lottery-type, gambling tasks. This research supports the proposition
that individual utility functions can be derived for some decision tasks. More
importantly, these studies support the basic underlying assumption of SEU theory;
individuals can and do make decisions which attempt to maximize utility, at least for

the tasks studied.

11
WW DCSPite the body of

evidence cited above, others have argued against the usefulness and accuracy of
subjective expected utility theory. A primary criticism leveled against SEU is that the
theory cannot be disproved given its normative nature. McCord and de Neufville
(1983), for example, argued that empirical evidence suggesting that individuals do not
attempt to maximize expected utilities in their decisions can be dismissed on the
grounds that these individuals are not behaving rationally. Proponents of SEU may
simply argue that these individuals are not cognizant of their irrational decision
behavior and that once aware, they would behave rationally and attempt to maximize
utility.

Given the rationality predicament, the common approach taken to refute SEU
theory has been to disprove the individual axioms upon which it is based (e.g.,
Hagen, 1979; MacCrimmon & Larsson, 1979; Kahneman & Tversky, 1979).
Compared to the other four axioms, by far the most damaging empirical evidence has
been leveled against the transitivity or the complete ordering axiom. According to
this axiom, if an individual prefers A to B and B to C, then A should also be
preferred to C. Given equal expected values (i.e. , equivalent attractiveness) and
probabilities, the order of preferences is expected to hold across lotteries. AS
demonstrated by the considerable empirical research generated by prospect theory
(Kahneman & Tversky, 1979), this is often not the case. Prospect theory offers an

alternative theoretical perspective on individual risky decision making which includes

12
propositions that contradict those generated by SEU theory. Prospect theory, its basic

tenets, and empirical research generated by the viewpoint are discussed below.
W
Kahneman and Tversky (1979) developed prospect theory as an alternative to
subjective expected utility theory, contending that the latter theory did not offer an
accurate description of individual risky decision making. Prospect theory attempts to
account for the nonrational violations of SEU theory that are apparent in individual
decision making. At a fundamental level, prospect theory is based on three tenets or
propositions which violate the axioms of the SEU model. These three tenets, labelled
the certainty, reflection, and reference effects, are described below.
W. The certainty effect is the contention that individuals
overweight outcomes that are considered certain relative to outcomes which are
considered probable, when choosing between two positive potential outcomes. The
certainty effect, also known as the Allais paradox, is illustrated by examining the
pattern of preferences generated by the following two choice problems borrowed from
Kahneman and Tversky (1979):
Problem 1: Choose between options A and B:
A. $4,000 with probability .80
B. $3,000 with certainty
Problem 2: Choose between options C and D:
C. $4,000 with probability .20

D. $3,000 with probability .25

13
In Problem 1, 80 percent of the subjects selected Option B and 20 percent

chose Option A. In terms of the subjective expected utility model, this implies that
u(3,000)/u(4,000) > 4/5, where ”u” denotes the utility of the outcome. However in
Problem 2, 65 percent of respondents selected Option C and 35 percent chose Option
D, implying the reverse inequality (i.e., u(3,000)/ u(4,000) > 4/5). Notice also, that
Option C (4,000, .20) can be expressed as 25 percent of Option A (A, .25).
Likewise, Option D (3,000, .25) can also be considered 25 percent of Option B (B,
.25). According to subjective expected utility theory, if B is preferred to A, then (B,
.25) ought to be preferred over (A, .25). However, 65 percent of Kahneman and
Tversky’s subjects violated the SEU prediction by selecting Option C (A, .25). It
seems that reducing the probability of winning from 1.0 to .25 has a greater effect
than a reduction from .80 to .20. This is not a finding that can be accounted for by
subjective expected utility theory.
Kahneman and Tversky (1979) have also demonstrated the certainty effect
using non-monetary outcomes. Consider the following choice problems:
Problem 3: Choose between options A and B:
A. 50 percent chance to win a three-week tour of England, France,
and Italy
B. A one-week tour of England, with certainty
Problem 4: Choose between options C and D:
C. 5 percent chance to win a three-week tour of England, France,

and Italy

14

D. 10 percent chance to win a one-week tour of England

In Problem 3, Option A was selected by 22 percent of the subjects and Option
B was chosen by 78 percent. In Problem 4, however, 67 percent of the subjects
chose Option C while 33 percent selected Option D. In this problem set, Option C
can be expressed as (A, .10) and Option D can be rewritten as (B, .10). Nonetheless,
given the strong preference for Option C it seems that a reduction in probability from
1.0 to .10 has a greater effect than the reduction of .50 to .05.

The strong tendency for individuals to overweight certainty relative to
probability is not adequately captured by subjective expected utility theory. Prospect
theory, however, does incorporate the certainty effect. The certainty effect is an
integral aspect of the prospect theory model, which also helps describe and account
for the second tenet of prospect theory, the reflection effect.

W. While the certainty effect is concerned solely with
positive prospects, the reflection effect describes decision behavior related to both
positive and negative prospects. The reflection effect describes differences in decision
making that are attributable to the way in which a choice problem is presented or the
manner in which it is "framed" (Tversky & Kahneman, 1981). Kahneman and
Tversky’s (1979) reflection effect deals with choice problems presented or framed as
gains and those presented or framed as losses. Their contention is that preference
patterns for negatively framed decision alternatives (i.e. , those framed as losses) are
the mirror image (i.e. , the reflection) of preference patterns for positively framed

choices (i.e., those framed as gains). The following example, taken from Kahneman

15

and Tversky (1979), demonstrates the difference between gain and loss frames and
illustrates the reflection effect:
Problem 5: Choose between options A and B:
A. Gain $3,000 with probability .90
B. Gain $6,000 with probability .45
Problem 6: Choose between options C and D:
C. Lose $3,000 with probability .90
D. Lose $6,000 with probability .45
Confronted with Problem 5 which is framed positively as a gain, 86 percent of
the subjects selected Option A and 14 percent preferred Option B. The preference
pattern for Problem 6, however, is reflected such that 8 percent of the subjects chose
Option C and 92 percent selected Option D. Problem 6 is framed negatively as a
loss. These results demonstrate the major implication of the reflection effect:
individuals are risk averse in the positive domain (i.e., under gain frames) and risk
seeking in the negative domain (i.e., under loss frames). This contention is counter
to subjective expected utility theory axioms. According to SEU theory, the frame
should be irrelevant and the ordering of preferences should not be reflected, but
remain consistent.
The reflection effect has also been demonstrated with non-monetary choice
problems. Of these decision tasks, the "Asian disease” problem and the results
associated with it are probably the most widely known (Tversky & Kahneman, 1981).

This decision problem is presented below:

16
Problem 7: Imagine that the US is preparing for the outbreak of an unusual

Asian disease, which is expected to kill 600 people. Two alternative programs
to combat the disease have been proposed. Assume that the exact scientific
estimate of the consequences of the programs are as follows:
A. If Program A is adopted, 200 people will be saved.
B. If Program B is adopted, there is 1/3 probability that 600 people
will be saved, and 2/3 probability that no people will be saved.
Problem 8: Same as Problem 7 with the following choice alternatives:
C. If Program C is adopted 400 people will die.
D. If Program D is adopted there is 1/3 probability that nobody
will die, and 2/3 probability that 600 people will die.
Considering the two decision problems together, it is clear that Programs A
and C offer equivalent outcomes, as do Programs B and D. Yet, in Problem 7 the
majority of individuals (72 percent) selected Program A, while the majority (78
percent) chose Program D in Problem 8. As was the case in the monetary decision
task, these results are attributed to differences in frame, a gain frame in Problem 7
and a loss frame in Problem 8. Similar results have been obtained using gambling
and purchasing decisions (Tversky & Kahneman, 1981). Given the consistency of
these results, Kahneman and Tversky (1984) have argued that framing is an extremely
Strong determinant of choice, stating that, ”the failure of invariance is both pervasive

and robust. It is as common among sophisticated respondents as among naive ones,

17

and it is not eliminated even when the same respondents answer both questions within
a few minutes” (p. 343).

The certainty effect, or the tendency to prefer guaranteed outcomes, is a
phenomenon exclusive to the gain domain. However, according to Kahneman and
Tversky (1979), the psychological phenomenon of overweighting certainty helps
account for the differential preferences in gain and loss domains. Operating under
gain frames, individuals are risk averse such that they prefer a smaller certain gain
over a larger one that is probabilistic. Under loss frames, individuals are risk taking
such that risking a greater loss is preferable to accepting a smaller certain loss. For
gain frames certainty is overweighted in a positive manner, making small certain f
gains more attractive. For loss frames, certainty is overweighted in a negative
manner, such that small certain losses are more aversive. Thus, overweighting
certainty in both a positive and negative manner in association with gain and loss
frames respectively, offers a partial explanation for the reflection effect. The
reflection effect may also be understood in terms of the reference effect, described in
the next section.

W. According to prospect theory, decision alternatives are
not evaluated in terms of final outcomes yielded (e. g., the ultimate amount of wealth
acquired). Rather, Kahneman and Tversky (1979) argued that decision outcomes are
evaluated as either gains or losses relative to some neutral reference point. The status

of an individual’s current assets is most often considered the neutral reference point,

18
with the understanding that the reference point may be affected by expectations of the

decision maker and the formulation of decision alternatives.
The effect of a neutral reference point based on one’S current assets is
illustrated in the following example borrowed from Kahneman and Tversky (1979):
Problem 9: In addition to whatever you own, you have been given $1 ,000.
You are now asked to choose between A and B.
A. Gain $1,000 with probability .50
B. Gain $500 with certainty
Problem 10: In addition to whatever you own, you have been given $2,000.
You are now asked to choose between C and D.
C. Lose $1,000 with probability .50
D. Lose $500 with certainty
Notice that the two problems are equivalent in terms of final states. Selecting
Option A or C affords a 50 percent chance to end up with $2,000 and a 50 percent
chance to finish with $1,000. Choosing Option B or D provides $1,500 with
certainty. Despite equivalent outcomes, the pattern of preference differed between the
two problems. The majority of subjects confronted with Problem 9 preferred Option
B (84 percent), while only 16 percent selected Option A. In Problem 10, 69 percent
of the subjects selected Option C and 31 percent chose Option D.
This problem demonstrates that decisions are not made on the basis of final
outcomes. Instead, decision alternatives are evaluated in terms of changes in

"wealth," either gains or losses. Change is evaluated on the basis of a neutral

19

reference point which is most typically determined by current asset levels. For
example, it seems that $1,000 was the reference point in Problem 9, while $2,000
was the reference point in Problem 10. Thus, gains and losses are not objective with
respect to some final asset state. Instead, whether a decision alternative is perceived
as a gain or a loss depends upon an individual’s reference point. This is an important
distinction between subjective expected utility theory and prospect theory. SEU
theory contends that utility is assessed and decisions are made on the basis of final
outcomes. From the prospect theory perspective, changes in outcome level are
clearly more important than the ultimate level of outcomes.

W. The certainty, reflection, and
reference effects, discussed above are useful in describing the prospect theory model
of the decision making process. Kahneman and Tversky (1979) suggest that decision
making consists of two major phases: the editing phase and the evaluation phase. In
the editing phase, prospects are interpreted as gains or losses in accordance with the
reference effect. That is, an individual evaluates whether a decision alternative
represents a gain or a loss as compared to a neutral reference point, usually the
individual’s current asset level.

In the evaluation phase, the individual attempts to assess the value of various
prospects or decision alternatives. Determining the highest value is not a
Straightforward task, according the Kahneman and Tversky (1979), because
individuals are subject to the biases associated with the certainty and reflection

effects. In accordance with the certainty effect, decision alternatives that provide

20

guaranteed outcomes are overweighted. In addition, the manner in which this
overweighting is manifested is specified by the reflection effect. In the positive
domain when gain frames are operating, certainty is overweighted such that
guaranteed smaller gains are evaluated much more favorably than probabilistic larger
gains. In the negative domain when loss frames are operating, certainty is
overweighted such that a sure smaller loss is viewed as more aversive than a
probabilistic larger loss.

The typical results of the evaluation phase can be illustrated using the
hypothetical value function Shown in Figure 1. According to Kahneman and Tversky
(1979), ”the value function is (a) defined on deviations from the reference point; (b)
generally concave for gains and commonly convex for losses; (c) steeper for losses
than for gains” (p. 279). As the final Step in the evaluation phase, prospect theory
contends that outcome values are multiplied by decision weights. Decision weights
are not necessarily the same as probabilities. Kahneman and Tversky argue that a
decision weight is likely to be a function of Stated probabilities, but that decision
weights may also be influenced by other factors (e. g., ambiguity). An individual
ultimately selects the decision alternative associated with the largest product, once
values and decision weights are multiplied.

W. Much of the supportive empirical research related to
prospect theory has already been described in this proposal. Nonetheless, it is worth
summarizing some of the key characteristics of that research. First, all of the

decision problems are hypothetical situations presented in written form. Second,

21

VALUE

 

GAINS

 

 

LOSSES

 

Figure 1. A hypothetical value function.

 

22

reflection decision problems are paired such that one is framed as a loss and its match
is framed as a gain. Although the majority of research was designed such that
different groups of subjects responded to only one problem in the pair, results
consistent with prospect theory predictions were still obtained when a Single group of
subjects responded to both the loss and gain versions of a problem (Kahneman &
Tversky, 1984). “Third, the majority of subjects employed in these studies were
students. Finally, the vast majority of empirical research supportive of prospect
theory has been conducted by Kahneman and Tversky, the theory’s founders. In each
of their studies, individuals presented with a positively fiamed choice problem made
risk averse decisions, while individuals confronted with a negatively fiamed choice
problem made risk seeking decisions.

This is not to say that all empirical research related to prospect theory has
been supportive. Using a cancer treatment decision problem, similar to Kahneman
and Tversky’s ”Asian disease" problem, Fagley and Miller (1987) found that subjects’
choices between risky and certain options were not significantly affected by decision
frame. This study is interesting because Fagley and Miller originally set out to
determine whether or not training in decision theory would mitigate the effects of loss
and gain frames on decision riskiness. Their intent was to examine the decision
patterns of 45 MBA students before and after training. However, Fagley and Miller
discovered that no framing effects were present even before training in decision
theory. Both before and after training, subjects generally preferred the certain

decision alternative in both the positively and negatively framed conditions. Thus,

 

23

their results were consistent with prospect theory for positively framed decision
problems, but not for those framed negatively.

Fagley and Miller’s (1987) research differed from Kahneman and Tversky’s
work in at least two critical ways. First, the subjects were MBA students enrolled in
an advanced statistical decision theory course. Fagley and Miller speculated that
perhaps these MBA students possessed some particular statistical knowledge that
affected the results. They also considered it possible that simply being enrolled in a
course on advanced decision theory sensitized subjects to be more aware of the
implications of the decision alternatives, especially since the problems were presented
during regular class time. However, these explanations seem improbable given
Kahneman and Tversky’s ( 1984) work demonstrating no framing effect differences
related to expertise. Finally, given the preference for certain alternatives
demonstrated in their study, Fagley and Miller also conjectured that their particular
group of MBA students may have been especially conservative. However, they were
unable to verify this notion.

The second critical distinction between Kahneman and Tversky’s work and
Fagley and Miller’s (1987) study was that subjects involved in the latter were asked to
explain the rationale for their choices. Asking individuals to explain the rationale is
equivalent to asking them to justify their decisions. It is possible that this group of
MBA students Simply felt more comfortable justifying a conservative choice. It is
also possible that asking individuals to focus on the decision rationale may have

significantly altered the decision process itself (Ericsson & Simon, 1980; 1984).

24
Thus, Fagley and Miller’s ( 1987) results may be attributable to important

characteristics of the subjects or the procedures employed.

In another study, Fagley and Miller (1990) attempted to study the effects of
sex and decision problem on the relationship between framing and risky decision
making. Based on prior research suggesting that females tend to be more field
dependent than males, Fagley and Miller argued that females may be more likely to
rely on frames in making decisions. Male and female undergraduate students were
employed as subjects. The five hypothetical decision problems employed were paper
and pencil measures taken from previous research and based on the following issues:
cancer treatment, job layoffs, dropout prevention, civil defense, and the Asian disease
scenario. There was a significant framing effect for females in the direction predicted
by prospect theory on four of the five decision problems (i.e. , all except for the
dropout prevention scenario). There were no Significant framing effects for males on
four of the decision problems. However, males’ decisions on the cancer treatment
problem were affected by gain and loss frames in the direction opposite of that
predicted by prospect theory. Males were risk averse under a loss frame and risk
seeking under a gain frame.

Even though results generally supported the frame by sex interaction
hypothesized by Fagley and Miller (1990), many of their findings are not clearly
explained. For instance, they were not able to specify which aspects of the dropout
prevention problem, if any, were responsible for the lack of framing effect for female

decision makers. Likewise, they were not able to explain the unexpected effects (i.e. ,

25

counter to prospect theory) obtained for male decision makers on the cancer treatment
problem. This finding is especially puzzling since Fagley and Miller (1987) found
that individuals confronted with the cancer treatment problem generally tended to opt
for the certain decision alternative, regardless of frame.

W. It is difficult to come to an overall
conclusion regarding prospect theory based on available empirical evidence. Using a
”scorecard" strategy to summarize the literature, it is clear that the majority of
empirical studies have supported prospect theory predictions (e.g. , Kahneman &
Tversky, 1979, 1984; Tversky & Kahneman, 1981), while only a handful have Shown
no effects or effects opposite to prospect theory predictions (e.g. , Fagley & Miller,
1987; 1990). Regardless, the empirical work related to prospect theory suffers from
two potential limitations.

The first concerns the primary way in which prospect theory has been tested,
using paper and pencil hypothetical choice problems. The problem is not inherent to
paper and pencil measures. The problem arises from testing a theory using primarily
as Single method, regardless of what that method is (Flatt, 1964.) Similar findings
obtained using multiple methods provide a more persuasive case in support of any
theory. There are a few prospect theory studies that have employed alternative
methods and those are discussed in later sections of this proposal (e.g., Kameda &
Davis, 1990; McGuire, Kiesler, & Siegel, 1987; Schurr, 1987). However, the vast
majority of studies have employed similar hypothetical choice problems, prompting

one to be cautious in expressing the generalizability of prospect theory.

26
The second problem with the empirical work related to prospect theory is

similar to the first. Just as Platt (1964) cautions against using a single method, he
also cautions against relying too heavily on the results of a Single researcher or group.
It is clear that Kahneman and Tversky’s empirical work has been consistently
supportive of prospect theory (e.g. , Kahneman & Tversky, 1979, 1984; Tversky &
Kahneman, 1981). Recently, Fagley and Miller and their colleagues have begun
generating research related to prospect theory and most of it has not been supportive
(e.g., Fagley & Miller, 1987; 1990). Other researchers have recently provided and
will undoubtedly continue to contribute additional studies, which will help alleviate
”pet theory" concerns (e.g., Kameda & Davis, 1990; McGuire et al., 1987; Schurr,
1987).

The most definitive statement that can be made regarding research on the
flaming effects predicted by prospect theory is that empirical findings have been
inconsistent. The present challenge is to discover those conditions under which
prospect theory predictions do fit the data and those circumstances under which
predictions do not fit the data. In essence, this means identifying those factors that
define the boundaries of prospect theory’s framing effects. AS an initial step in this
endeavor, it seems reasonable to explore how prospect theory can be reconciled with
the rational model of decision making, subjective expected utility theory.

l!‘ 4111’ _;_.-.!.I'. 1.1!. A 0 -_ ' . Mm '_ 0 he. 'i nan:
Attempting to determine which model of risky decision making is correct,

subjective expected utility theory or prospect theory, would be a futile task. Both

27

have considerable theoretical and empirical justification. Each offers a reasonable
description of risky decision making behavior under certain circumstances.

Therefore, it seems that neither theory could or should be easily dismissed. A more
fruitful approach to the study of risky decision making is to determine the ways in
which the two theories complement one another and to ascertain how they can be used
together to more fully understand risky decision making.

Subjective expected utility theory describes the rational normative model of
decision making. It also describes the optimal method for maximizing outcomes.
Generally, it seems reasonable to assume that individuals would like to maximize their
outcomes when possible. Nonetheless, individuals’ decision choices are not always
consistent with the normative model. Particular influences seem to pull individuals
away from the completely rational model. Prospect theory describes some of these
influences (e. g. , cognitive decision frames), allowing one to begin to determine the
circumstances under which nonrational decision making is most likely.

According to prospect theory, cognitive decision frames have a substantial
influence on risky decision making. Under gain frames, individuals tend to be risk
averse, preferring smaller certain gains over larger risky gains. Under loss frames
individuals are more risk seeking, chancing a larger loss to avoid a smaller certain
loss.

A primary goal of the present research is to explore some of the boundary
conditions of cognitive framing effects. Stated another way, I am concerned with

discovering conditions under which framing effects are most likely to occur and those

28

conditions under which framing effects are least likely to occur. To begin to
understand the former, the characteristics of decision tasks known to elicit framing
effects must be examined.

Recall that most prospect theory research, and all the original research by
Kahneman and Tversky (1979; 1981; 1984), employs a Similar type of decision task.
Participants are presented with a hypothetical scenario, one that is either framed as a
gain or framed as a loss, and asked to make a choice between a more risky (i.e.,
lower probability) option that offers greater potential outcomes and a certain
alternative (i.e. , the associated outcome is guaranteed) that offers a relatively lower
outcome. Prospect theory predicts that under gain frames, individuals will be more
likely to opt for the certain alternative that provides less of a given outcome than the
risky alternative that has the potential to yield a larger outcome. Under loss frames,
prospect theory makes the opposite prediction. Individuals are expected to choose the
risky outcome-maximizing alterative over the certain alternative.

The following pair of decision problems, taken from Kahneman and Tversky
(1979), may help clarify prospect theory predictions. The sample problems may also
be useful for demonstrating how prospect theory predictions differ from those of
subjective expected utility theory (SEU):

Gain Frame: In addition to whatever you own, you have been given $1 ,000.

You are now asked to choose between A and B.

A. Gain $1,000 with probability .50

B. Gain $500 with certainty

29

Loss Frame: In addition to whatever you own, you have been given $2,000.
You are now asked to choose between C and D.

C. Lose $1,000 with probability .50

D. Lose $500 with certainty

Recall that in research employing the above problems (Kahneman & Tversky,
1979), the majority of subjects in the gain frame condition selected Option B, the
certain alternative. Under the loss frame, the majority chose Option C, the risky
alternative. These findings are consistent with prospect theory.

Subjective expected utility theory, in particular the transitivity axiom, suggests
that individuals will make the same type of outcome maximizing choice regardless of
decision frame. Note that under both gain and loss frames, the risky alternative is
always associated with a larger gain or a smaller loss. Since SEU posits that
individuals behaving rationally will attempt to maximize outcomes, the theory would
most likely predict that individuals will generally prefer Options A and C (i.e. , those
associated with higher outcomes) to Options B and D in the present example.

It is important to recognize that under loss frames both SEU and prospect
theory predict the same choice, the election of the outcome-maximizing risky
alternative (i.e. , Option C in the example above). For gain frames, however, SEU
predicts that decision making will be consistent with loss frame choices (i.e. , the
riskier outcome-maximizing option will be preferred), while prospect theory proposes

a differential preference for certain alternatives. In terms of the decision problem

30
presented above, SEU predicts that Option A will be chosen while prospect theory

predicts that Option B will be chosen.

In an effort to reconcile SEU and prospect theory it may be useful to focus
first on decision making under a gain frame, the condition under which the two
theories actually make differential predictions. The primary goal is identifying some
boundary defining process or condition that helps us understand when the predictions
of SEU are more likely to be accurate than the predictions of prospect theory and vice
versa. After a discussion of gain frames, it may then be helpful to focus on loss
frames in an attempt to decipher the psychological processes that may account for
frame-dependent differential decision making.

mm. AS mentioned previously, a major limitation of prospect theory
research is the use of a single method, decision tasks that are predominantly paper
and pencil, hypothetical, choice problems. These decision scenarios are essentially
lotteries for which outcome probabilities are provided. By definition, a lottery is a
chance event. Chance, by definition, possesses a degree of unpredictability and a lack
of control. The sample decision problems provided above help clarify the
significance of these definitions.

If an individual selects Option B, described earlier, he or She will definitely
acquire $500. However, if the individual selects Option A there is a 50 percent
chance that he or she will gain $1,000 and a 50 percent chance that nothing will be
gained. Given that the stated probabilities are 50/50, whether the individual falls into

the former or the latter category on any one trial is equally probable. Since the

31

decision problem is a lottery, there is little reason for an individual to have greater
confidence in one outcome over the other. The individual has no personal control and
no reason to believe that he or she has any influence over the probabilities. Thus,
instead of leaving a gain up to chance the individual can, and does according to most
prospect theory research, ensure a gain by selecting Option B (i.e. , the risk averse
choice), even though the certain gain is smaller.

Consider a hypothetical scenario. An individual is confronted with a choice
between Options A and B. If that person selects Option A, he or she must make a
correct decision in order to receive the $1,000. AS stated in Option A, the
probability of making a correct decision is .50. If the individual chooses Option B,
he or she will automatically receive $500 without nwding to make a correct decision.
The argument made here is that in this context, whether an individual chooses Option
A or B is dependent upon that person’s W
W. The stated probability for making an accurate decision is
.50. However, the W may be something quite different. For
instance, the individual may have any number of reasons to feel absolutely certain that
he or She will answer the question correctly (e. g. , the person is highly
knowledgeable, very experienced, etc.). In that case, the individual’s perceived
probability would be 1.0. The stated probability is disregarded because the individual
believes that he or she has influence or control over the probability of being correct.
Given a choice between Option A and B under these circumstances, there can be no

doubt that an individual would select Option A because it has as higher pay-off.

32

Here it is argued that the decision maker’s perceived control over the
probability of making a correct decision is a critical moderator of the effects of
framing on risky decision making. Perceived control may be a critical boundary
condition that plays a role in determine whether prospect theory predictions or SEU
predictions are more likely to be accurate under gain frames. It is expected that the
greater the control perceived by the decision maker, the greater the likelihood that the
risky alternative (i.e., the outcome maximizing option) will be chosen and SEU
predictions will be supported. The less control perceived or the more the decision
task appears to be a lottery, the greater the likelihood that the certain alternative (i.e. ,
the smaller guaranteed outcome) will be chosen and prospect theory predictions will
be supported.

mm. The notion of control is useful for understanding risk averse
decision maldng under gain frames, but does not explain the risk seeking decisions
made under loss frames. As typically presented, prospect theory scenarios, both
those presented as gains and those presented as losses, represent low control
situations. Thus, individuals do not select riskier alternatives because they perceive
greater control due to the loss frame. To understand decision behavior under loss
frames, one must understand the psychological meaning of the loss condition.

Essentially, potential losses are perceived as threatening. Jackson and Dutton
(1988) conducted a study designed to determine the critical features that distinguish
opportunities from threats. A large sample of M.B.A. alumni were asked to read

several business strategy scenarios and rate them on the extent to which they

33
represented threats or opportunities. Jackson and Dutton found that low control,

potential for loss, and no potential for gain were critical features distinguishing threats
from opportunities. Thus, a loss frame in a prospect theory scenario meets the
definitional criteria of a threat.

Individuals are expected to respond differently to threats than to opportunities.
Dutton and Jackson (1987) argued that managers dealing with Strategic issues would
take risks of greater magnitude in response to threatening conditions than in response
to opportunities. Consistent with prospect theory predictions, managers were
expected to be willing to take greater risks to avoid loss than to obtain gains. The
present proposal contends that the influence of loss aversion is so great that it is
present regardless of control. Under loss frames, individuals are expected to be risk
seeking regardless of the level of perceived control.

W

Perceived control may be a critical boundary condition of the effects of
cognitive frames, specifically gain frames, on risky decision making. The concept of
perceived control iS not new to the psychological and organizational literatures. It has
been a Significant and prevalent construct in psychological thought for nearly 100
years (e.g., Gross, 1901). In order to more fully understand the concept of perceived
control and its potential relevance for risky decision making, some of the more
pertinent literature is discussed below.

.' a! Air. '1 e “IVA. 1.1!0‘«_J...2|0n-_ {111.1}. Sutton

and Kahn (1986) contended that the notion of perceived control is a particularly

34

important construct in the organizational literature. The primary argument in the
literature is that control is desirable and human beings are motivated to achieve and
maintain a perception of control in the environments of which they are a part
(Greenberger & Strasser, 1986; Rothbaum, Weisz, & Snyder, 1982). In general,
perceptions of greater control mean that individuals feel a sense of mastery and
personal competence in their environments (deCharms, 1968; White, 1959). Lower
perceived control, on the other hand, creates feelings of uncertainty and lack of
mastery that are undesirable and aversive (deCharms, 1968; Greenberger & Strasser,
1986). Perceived control has been linked to satisfaction (Greenberger, Strasser,
Cummings & Dunham, 1989; Tetrick & LaRocco, 1987), performance (Bazerrnan,
1982; Glass & Singer, 1972; Greenberger et al., 1989), stress (Averill, 1973; Miller,
1977; Tetrick & LaRocco, 1987; Thompson, 1981) and withdrawal (Langer & Rodin,
1976).

Spector (1986) conducted a meta-analysis of the effects of perceived control on
a variety of outcome variables in the work context. According to Spector’s results,
perceived control was positively related to job satisfaction, organizational
commitment, job involvement, performance, and motivation. In addition, perceived
control was negatively related to stress, absenteeism, intentions to turnover, and
turnover.

W. A much more limited literature addresses
the relationship between perceived control and risk taking. This literature is entirely

theoretical and offers no empirical data to support hypothesized relationships between

35
perceived control and risk taking. Vlek and Stallen (1980) developed a conceptual

model of the determinants of risk taking. They contended that perceived control
would be positively related to risk taking (i.e. , the greater the perceived control the
greater the risk taking demonstrated). In particular, they argued that risk taking
would be greater for skill-dependent decision tasks than for chance-dependent tasks.
Baird and Thomas (1985) extended and applied this notion to a conceptual model of
strategic risk taking for managers. They asserted that controllability of decision
consequences would have a positive impact on the likelihood of accepting a high
strategic risk. These examples from the risk literature converge to support the
theoretical notion that control is positively related to risk taking.

The unique contribution of the present proposal is applying the notion of
perceived control in an attempt to explain the risk averse decision making frequently
evidenced under gain frames. The individual decision maker’s tendency to select
more certain alternatives under gain frames is hypothesized to be a function of the
low perceived control, fostered by the lottery-type tasks used in prospect theory
research. Control is proposed as a moderator that begins to define the boundaries of
prospect theory’s framing effect under gain conditions.

WWW Since the present research
focuses on risky decision making, perceived control is conceptualized in terms of the
decision maker’s perceived probability of making a correct decision. The most
simple instance of a risky decision involves a choice between two alternatives where

one option is correct and the other incorrect, and the lowest probability of making a

36

correct decision is .50. Any individual has at least a 50 percent chance of making a
correct decision. The highest probability of making a correct decision is 1.0 which
means the correct decision will be made with certainty. An individual’s perceived
probability can range from .50 to 1.0.

The perceived probability of making a correct decision may be, but does not
necessarily have to be, consistent with the actual or objective probability of making a
correct decision. Furthermore, perceived probabilities may not be consistent with
stated or average probabilities for a given decision task. A substantial amount of
research, most of it conducted under the rubric of "unrealistic optimism, " indicates
that individuals often perceive personally applicable probabilities as more favorable
than Stated or average probabilities across a variety of issues and contexts (e.g. ,
Langer & Roth, 1975; Weinstein, 1980). The decision making literature also Shows
that individuals tend to believe they will make more accurate decisions than they
actually do (Lichtenstein, Fischhoff, & Phillips, 1982; Slovic, Fischhoff, &
Lichtenstein, 1982).

To explore the proposed moderating effect of perceived control on the
relationship between decision frame and decision riskiness, it becomes important to
utilize non-lottery tasks. As discussed above, the typical prospect theory decision
scenario is unlikely to foster feelings of control. The effects of perceived control may
be more relevant and obvious within a complex decision making environment which
actually engages the decision maker. It seems, for instance, that information

processing-type tasks may be more likely to provide the requisite opportunities for a

37

decision maker’s perceptions of control to come into play. These kinds of tasks
require an individual to measure, process, evaluate, and combine informational cues
to reach a decision. Unlike lottery tasks, information processing tasks can have
qualitatively correct responses. In addition, the results obtained using information
processing tasks are likely to have greater generalizability than the written
hypothetical choice problems most frequently employed.

Most importantly, information processing tasks can be manipulated to be more
or less like a lottery task. Using information processing tasks, variance on critical
task characteristics can be achieved. For instance, the informational decision cues can
vary in terms of ambiguity (i.e., the extent to which information provided is clear or
unclear). Unambiguous cues are expected to make the correct decision more obvious
and give the decision maker greater control. It is assumed that highly ambiguous
cues reduce personal control because they provide little discriminating information.
Thus, a decision maker confronted with ambiguous cues is likely to perceive less
control because the lack of clarity diminishes his or her ability to make a correct
decision.

Information processing tasks also allow decision makers to assess certain types
of information that are likely to influence perceptions of control over the probability
of making a correct decision. By definition, people cannot determine their expertise
on lotteries. At least theoretically, one’s previous performance history in a lottery
has no bearing on current or future performance. However, an individual would be

able to ascertain his or her expertise and performance history on an information

38

processing task. Since this type of information may have a direct bearing on
perceived levels of control, its availability creates the opportunity to test the predicted
moderating effect of control on the relationship between decision frame and decision
riskiness.

Finally, it is important to be explicit about how perceived control will be
manifested in an information processing context for this study. AS mentioned above,
the type of information processing task used in the present research requires a
decision maker to measure, process, evaluate, and combine informational cues in an
attempt to reach an objectively correct decision. AS with the hypothetical lottery
tasks typically employed in prospect theory’s cognitive framing research, decision
alternatives and the potential outcomes associated with each alternative are provided.
The certain alternatives have guaranmd outcomes. The risky alternatives offer
greater pay-offs, provided that a correct decision is made. Individuals perceiving
greater control are predicted to make riskier decisions. In this context, individuals’
perceptions of control will be operationalized as the decision maker’s perceived
probability for making a correct decision. That is, the probability that the decision
maker mugs, not necessarily the stated or accurate probability. Thus, a
decision maker who has greater perceived control believes the probability of making a
correct decision is higher than one who perceives less control.

W191. The present research examines perceived
control as a potential moderator of the relationship between decision frame and

decision riskiness. The nature of the proposed moderator effect iS shown in Figure 2.

39

High Control

 

High

RISK

Low Control

Low

 

 

Gain Loss
DECISION FRAME

Figure 2: Hypothesized moderating effect of control

40

Perceived control is expected to lead to differential decision making under gain
frames. Under a gain frame, an individual working within conditions of low control
(i.e. , typical prospect dreary conditions) is predicted to make risk averse decisions.
As shown in the graph, decision riskiness increases as control increases. Notice that
under loss frames, individuals are expected to consistently make decisions reflective
of risk seeking. Prospect theory research representing conditions of low control has
shown that individuals are risk seeking under loss frames. Here it has been argued
that the psychological threat produced by loss frames is likely to engender risk
seeking regardless of the level of perceived control. The magnitude of perceived
control’s moderator effect, however, is a more difficult to predict. It is possible that
the line representing high control in Figure 2, may actually be more appropriately
drawn higher on the y-axis.
. .11 It; ' .L .3 I ‘1' ° Masn° 0‘ i-‘v F ‘ -. c. 1? glva! o I r

In summary, subjective expected utility theory describes a rational model of
decision making based on maximizing outcomes. Prospect theory invokes the notion
of cognitive decision frames to explain the deviations from rationality often observed
in individual risky decision making. Positive decision frames, in particular, are
associated with a preference for smaller certain gains in lieu of larger probabilistic
gains.

It is suggested here, however, that this type of framing effect may be limited
to the conditions of low perceived control fostered by characteristics of prospect

theory decision tasks. These tasks cannot be manipulated and offer no critical

41

discriminating information which would lead one to rely on perceptions of personal
control. When discriminating information (e.g. , cue ambiguity, task knowledge,
performance history) is available and individuals are able to assess their own feelings
of control (i.e. , probability of making a correct decision), decision frames are less
likely to have an effect on risky decision making. The potential moderating effect of
perceived control on the relationship between framing and risky decision making
would be best studied using more engaging information processing decision tasks.
The section that follows is devoted to the development of a model of individual
risky decision making. The moderating effect of perceived control on the relationship
between cognitive decision frame and decision riskiness forms the basis of the model.
The personal and informational characteristics likely to influence perceptions of
control are also considered. The section concludes with a description of individual

level hypotheses derived from the model.

EFFECTS OF COGNITIVE FRAMES AND PERCEIVED CONTROL
ON RISKY DECISION MAKING FOR INDIVIDUALS

The model presented in Figure 3, illustrates the linkages to be discussed in the
sections that follow. Keep in mind that the present research represents an initial
attempt to define and test the boundary conditions of prospect theory framing effects
(Sitkin & Pablo, 1992). As such, the proposed model Should be considered a
heuristic designed to direct the exploration of framing boundaries, rather than a
comprehensive illustration of the process. The role of perceived control as a
moderator of the relationship between framing and risky decision making was
extensively discussed in the previous section. A brief summary of the nature of the
moderating relationship and the literature supporting it is provided below. After that,
the specific ways in which the task and individual difference variables are expected to
affect perceptions of control are detailed.
BMW

Recall that the nature of the moderating effect of perceived control on the
relationship between decision frame and decision riskiness was shown previously in
Figure 2. Under gain frames, perceptions of control are expected to influence risk
taking. Here it is argued that decision Situations fostering perceptions of low control
are equivalent to chance lotteries. Under these circumstances, individuals operating
under gain frames are likely to opt for a smaller guaranteed gain. Under conditions

of high perceived control, decision makers operating under gain frames are more

42

43

 

LOCUS OF CONTROL
TASK KNOWLEDGE

PERFORMANCE HISTORY
CUE AMBIGUITY

 

 

 

 

PERCEIVED CONTROL

 

 

 

 

 

 

 

DECISION FRAME ‘r—JL-D DECISION RISK

 

 

 

 

 

Figure 3: Model of individual risky decision making

44

likely to select risky alternatives to maximize outcomes. This argument has been
presented elsewhere in the risk taking literature (e.g., Baird & Thomas, 1985; Vlek &
Stallen, 1980).

Under loss frames, individuals are predicted to make more consistently risky
decisions. Loss frames represent threatening situations (Jackson & Dutton, 1988).
Individuals are expected to take greater risks to avoid losses than they are to secure
gains (Dutton & Jackson, 1987). Prospect theory research representing conditions of
low control has Shown that individuals are risk seeking under loss frames. Here it is
argued that individuals operating under loss frames will also be risk seeking under
conditions of high control.

F ' r ' l

Perceived control has been advanced as potential boundary condition limiting
the effects of cognitive frames on risky decision making. This section is devoted to
an initial attempt at exploring some of the potential influences on perceived control.
Three possible sources of control are introduced. Consistent with general models of
perceptions, it is argued that perceptions of control are influenced, in part, by
characteristics of the perceiver (i.e. , the decision maker), the task, and the perceiver’s
history of interaction with the task. Within each of these three domains, one or more
key variables have been selected from the literature based on the contention that they
represent some of the more likely sources of influence on perceived control in the
Situation to be studied. Locus of control is the perceiver characteristic. The task

variable concerns the task’s predictability as represented by the ambiguity of

45

informational cues associated with the decision (i.e. , cue ambiguity). Two variables
stemming from the perceiver’s interaction with the decision task will be considered.
The first is the perceiver’s knowledge about the task, including the meaning of
relevant information and appropriate methods for incorporating the information to
make a decision. The second is the perceiver’s past performance history on the
decision task. Each of these key variables is discussed below.

mm. The concept of locus of control originated in Rotter’s
(1966) work on the effects of rewards on behavior. According to Rotter’s theory, an
individual’s locus of control can be captured by a continuum ranging from internal to
external. Individuals closer to the internal end of the scale perceive that events and
outcomes are contingent upon their own personal behavior. In contrast, individuals
characterized as more external perceive outcomes as contingent upon forces outside
themselves (e.g., chance, luck, other individuals). Thus, locus of control is a
personality characteristic representing the tendency to attribute causality either to
oneself (i.e., internal) or to the external environment (i.e., external). In the present
study, locus of control is expected to influence one’s perceived control or one’s
perceived probability of making a correct decision. Because they have a tendency to
believe that events are under their personal control, individuals who are more internal
are predicted to perceive greater control over decision tasks than those who are more
external. Internals are likely to think it more probable that they will make a correct
decision regardless of the actual probability of making a correct decision. They are

more likely than extemals to see themselves as personally capable (Andrisani &

46
Nestel, 1976). Extemals are likely to view the probability of correctly making a

risky decision as contingent upon external forces that are beyond their personal
control, very much like lotteries. Extemals are predicted to believe making an
accurate decision less probable.

W. Task knowledge in the decision making context of interest
refers to the decision maker’s level of expertise on the information processing task.
Concentrating on information processing tasks, expertise or task knowledge would be
reflected in the extent to which the meanings of informational cues are understood,
the extent to which relationships among cues are understood, and the extent to which
the individual understands how the above information should be combined and utilized
to make decisions. In terms of the risky decision making model, individuals with
greater task knowledge are expected to experience greater perceived control over the
probability of making a correct decision than individuals with low task knowledge.
The more knowledgeable a person is regarding decision relevant information cues, the
more likely that person is to feel a correct decision can be made (assuming that the
person believes he or she is knowledgeable).

Baird and Thomas (1985) considered the effect of knowledge in their
conceptual model of strategic risk taking. They argued that a decision maker’s
knowledge would have a positive impact on that individual’s confidence. They
further theorized that greater knowledge and confidence would result in greater

strategic risk taking. Here the prediction is Similar. High task knowledge is

47
predicted to lead to increased perceived control, which ultimately results in greater

risk taking.

W. Performance history refers to an individual’s pattern of
results or outcomes resulting from previous decision making on a particular task. An
individual’s performance history captures all previous choice-outcome relationships.
In order for performance history to impact upon an individual’s perceived control for
a subsequent decision, the individual must be aware of the performance history. That
is, previous decisions must have provided the individual with feedback.

Positive feedback is strongly related to feelings of personal control (Averill,
1973). Furthermore, a positive performance history which demonstrates that previous
choices have been correct, encourages future risk taking in order to obtain higher
outcomes (Sitkin & Pablo, 1992). A similar effect has been demonstrated with
gambling tasks involving actual monetary outcomes (Thaler & Johnson, 1990).

Thaler and Johnson found that when their subjects’ risky gambles resulted in gains,
they were more likely to continue to take risks. The assumption of the present model
is that if a positive performance history increases risk taking for a lottery-type task, it
is even more likely to increase risk taking in an information processing task. In the
latter task, high performance is likely to be attributed to meaningful personal and
Situational sources, thereby increasing perceived control. Thus, a positive
performance history is expected to influence perceptions of control, ultimately leading
to increased risk taking. Conversely a negative performance history is proposed to

reduce feelings of control. Recall that the effect of low control on risk taking

48
depends upon the decision frame. Individuals working under gain frames are likely to

be more risk averse when perceived control is low. Individuals operating under loss
frames are likely to take risks regardless.

W. Cue ambiguity refers to the extent to which informational
cues are clear or unclear in terms of their implications for making a correct decision.
For instance, suppose the task is to decide whether or not prospective students Should
be accepted into graduate school. Assume that there are two relevant factors, grade
point average and letters of recommendation. Applicant A has a 4.0 grade point
average and glowing letters of recommendation. Applicant B has a 1.5 grade point
average and uncomplimentary letters of reference. Clearly, Applicant A should be
accepted and Applicant B Should be rejected. In each case, the informational cues are
unequivocal. Now consider Applicant C who has a 3.0 grade point average, one
superb letter of recommendation, and two mediocre letters. It is more difficult to
decide whether or not to admit Applicant C because the information is ambiguous.

Notice that the difficulty does not reside in the amount of information
available. Ambiguity may be high even when there is an ample quantity of
information. Ellsberg (1961) pointed out that ambiguous information reduces
confidence at least as much as a lack of information. A one can be ambiguous
because its individual meaning is unclear or because it conflicts with other
informational cues (Ellsberg, 1961). In the example above, for instance, Applicant
C’s 3.0 grade point average is ambiguous in itself because it is neither outstanding

nor exceptionally poor. Applicant C’s letters of recommendation are ambiguous

49

because they conflict with each other. Furthermore, when the informational cues are
combined they represent an ambiguous picture that makes decision making difficult.

In the present model of risky decision making, individuals faced with
ambiguous cues are expected to perceive less control than decision makers confronted
with unambiguous informational cues. Similar arguments have been made in the
context of risk taking. Vlek and Stallen (1980), for example, argued that individuals
are averse to ambiguity because it reduces control. Armelius (1979) examined the
effect of cue ambiguity on confidence in making correct decisions. Armelius found
that confidence in accurate decision making was a direct function of decision task
predictability. In this study, predictability was defined as the certainty of the
relationships among informational cues. Thus, the relationship between cue ambiguity
and perceived control proposed here has both theoretical and empirical support.
IndiyidrtaLlayeLExmtheses

The primary linkages in the risky decision making model depicted in Figure 2
have now been described. In addition, some potential influences on perceptions of
control have also been considered. Before turning to team context issues and risky
decision making, the relationships illustrated in Figure 2 will be summarized as
formal hypotheses.

The model of Figure 2 shows a direct relationship between cognitive decision
frame and decision riskiness. This linkage recognizes the main effect for framing,
proposed and supported in the prospect theory literature. Recall that the nature of

that effect is as follows:

H1:

50

Decision frames have a direct effect on decision riskiness. Under conditions
where alternatives are equal in subjective expected utility and unequal in risk,
individuals presented with a gain frame will be more likely to select a less
risky alternative than those presented with a loss frame. Under the same
conditions, individuals presented with a loss frame will be more likely to select
a riskier alternative than those presented with a gain frame.

Figure 2 depicts perceived control as a moderator of the relationship between

decision frame and decision riskiness. In general, greater perceived control is

expected to increase risk taking. Under conditions of high perceived control, decision

riskiness is not expected to differ between gain and loss frames. Under conditions of

low perceived control, prospect theory framing effects are predicted. That is,

decisions should be more risk averse under gain frames and more risk seeking under

loss frames.

H2:

Perceived control will moderate the relationship between decision frame and
decision riskiness. Specifically, risk seeking will be greater under loss frames
than under gain frames when perceptions of control are lower. When
perceptions of control are higher, there are no predicted differences for
decision riskiness between gain and loss frames.

Finally, several factors are predicted to influence of perceived control.

Individual and task characteristics are both included in the model.

H3:

Across decisions, individuals with a more internal locus of control, greater

task knowledge, and a more positive performance history will have higher

51

perceptions of control than individuals with a more external locus of control,
less task knowledge, and a more negative performance history. For Specific
decisions, individuals with less informational cue ambiguity will have higher

perceived control than individuals with greater informational cue ambiguity.

RISKY DECISION MAKING IN THE TEAM CONTEXT

Individual level risky decision making is an interesting and significant topic in
its own right. However, in the present Study the individual level model also serves as
the basis for considering risky decision making in the team context. As Stated in the
introduction, one of the major goals in developing a model of individual risky
decision making was to apply it to decision making in teams. This chapter is devoted
to that endeavor.

Ilfieam Versus Indin'dual Dmisign Malg'ng

There are some definitional issues that must be addressed before the
application of the individual model to risky decision making in the team context can
be considered. It is important to understand the way teams are defined here, and how
that definition differs from traditional definitions of ”small groups. " In order to
contemplate the effects of team context issues on risky decision making, an
understanding of the critical features that distinguish teams is essential. In the
sections that follow, teams are defined and the key aspects of decision making in
teams are described.

W. Much of the literature and research relevant to teams has been
presented under the label of ”small groups.” As treated here, however, teams and
groups are not synonymous. For present purposes, teams will be thought of as a
qualitatively different from small groups. Not all small groups meet the particular

criteria used here to define teams.

52

5 3
Small groups are typically defined as (a) a collective of two or more

individuals who, (b) interact with one another, and (c) are interdependent in some
way. McGrath (1984) has elaborated on this definition contending that small groups
in real world settings must exist in some time frame. That is, a small group has a
past, a present, and a future.

Teams Share the four above characteristics with small groups but can be
distinguished by several additional critical features. To begin, a key difference
between small groups and teams is that interdependence in teams typically takes the
form of differentiated and assigned roles (Dyer, 1984). That is, each team member
performs a unique and specific function within the team. Differentiated roles are not
a defining characteristic of small groups. Although small groups are also
interdependent, that interdependence is not captured by unique roles among members.

Two other key features of teams are related to the notion of differentiated
roles. Unlike small groups, teams exist expressly for some task oriented purpose
(Ilgen, Major, Hollenbeck, & Sego, in press). Furthermore, in attempting to fulfill
that purpose, team members have Shared goals and objectives. These goals are
usually explicit, and each team member typically has some level of awareness of these
goals (Ilgen et al., in press). A team’s task or purpose could be balancing a budget,
flying an aircraft, or playing a basketball game. The compatible goal may be to cut
the budget by ten percent, safely land the aircraft, or win the game. The
differentiated roles within a team can be useful in helping the team realize the goals

related to its purpose.

54

Coordinated behavior, a final distinguishing feature of teams, is also essential
in order for a team to attain its objectives (Fleishman & Zaccaro, 1992). Although
each team member performs Specific tasks in accordance with his or her defined role
within the team, individual members must also coordinate their performance on those
tasks or bring their individual efforts together in some way to fulfill a common
purpose (Fleishman & Zaccaro, 1992). This coordinated behavior is most
instrumental in the achievement of the team members’ shared goals.

Morgan, Glickman, Woodward, Blaiwes, and Salas (1986) provide a definition
that seems to capture the essence of a team succinctly. They state that teams are,
"distinguishable sets of two or more individuals who interact interdependently and
adaptively to achieve specified, shared, and valued objectives. " (p. 3) Their
definition will be adopted here.

31W. There are many varieties of
teams that fit the definitional criteria provided above. A primary distinction among
these teams, is the way decisions are made. For instance, in some teams each
member participates in all aspects of the decision process and each assumes equal
responsibility for the team’s decision, reaching a final decision by consensus. Juries
are a good example of this type of decision making team. Other teams use a
democratic decision making process. In these teams, every member has an equal
voice and decisions are reached by majority rule.

In hierarchical teams with distributed expertise, an individual decision maker,

the team leader, has decision making responsibility. Subordinate members may have

55

a voice in team decisions in that they may be able to make recommendations to the
leader, but it is the team leader who is ultimately responsible for making the team
decision. The distributed expertise distinction means that the knowledge (i.e. ,
expertise) relevant to decision making is disbursed across team members. Each team
member is an expert in some area of potential consequence to the decisions being
made; no individual team member is expert in all aspects. A flight crew is a good
example of this type of team. The flight captain is the team leader. He or she has
ultimate decision making responsibility. Other members of the team, however, have
critical areas of specialization and may contribute to the decisions that are made (e.g.,
the co-pilot, the navigational officer).

The present research is interested Specifically in the decision making of leaders
in hierarchical teams with distributed expertise. This type of team was chosen for
two primary reasons. First, hierarchical teams are representative of the teams making
critical decisions in a variety of applied settings (e. g., military command and control
teams, top management teams, flight crews, surgical teams). Second, hierarchical
teams represent the next logical step in a progression moving from the individual to
the team level. Even though decision making takes place within a team context, it is
still an individual, the team leader, who is actually making the decision. Since most
of the work on risky decision making has been conducted at the individual level, the
existing literature is more likely to apply to hierarchical teams than to other types of
teams in which decisions are reached through dramatically different processes (e. g. ,

democratic teams, consensus-seeking teams).

56

The brief description of hierarchical teams with distributed expertise is
provided as background for considering the applicability of existing literature to
leaders making decisions in that context. Further elaboration, however, will be
necessary in developing a model of risky decision making applicable to the team
context when there is distributed expertise. The requisite detail is provided in a later
section.

Empirical Research: Application to Hierarchical Teams

The literature related to risky decision making at the individual level was
reviewed in the first section of this proposal. The literature reviewed in this section
is focused on risky decision making in a team context. A primary goal of this review
iS to evaluate the relevance of existing theoretical and empirical work for
understanding the risky decision making of hierarchical team leaders. Two research
domains address risky decision making in a team context, risky shift and prospect
theory. Each literature is considered below.
main

Most often in the literature, "risky shift" refers to the phenomenon which
demonstrates that individuals advocate greater risk taking after participating in a
group discussion. A smaller subset of research compares the riskiness of decisions
made by individuals to the riskiness of a consensus decision made by the same group
of individuals. The study of risky shift began with Stoner (1961) who found that
business students opted for higher degrees of risk taking after participation in a group

discussion than they had previously when working in private. Similar findings have

57
been obtained using a variety of samples (see Clark, 1971; Pruitt, 1971; Vinokur,

1971b for comprehensive reviews).

Typically, subjects participating in risky Shift research read a scenario in
which they are asked to offer advice to a hypothetical person confronted by a choice
dilemma. The choice is usually between a safe less attractive alternative and an
uncertain more attractive alternative. The most widely used instrument in risky shift
research is the Choice Dilemmas Questionnaire (CDQ), a paper and pencil measure
consisting of 12 hypothetical situations like the one described above (Stoner, 1961).
Risky shift research as a whole has been criticized for the extensive and almost
exclusive use of the CDQ. While the lack of methodological variety has certainly
been a concern, the instrument itself has also been criticized for presenting all
scenarios in a positive frame (Bazerman, 1984).

law. The majority of risky shift research focuses specifically
on individual decisions before and after group participation. There has been
considerable debate in the literature over the most appropriate theoretical explanation
for the risky shift phenomenon. The major rationales advanced include: diffusion of
responsibility (e.g., Wallach, Kogan, & Bem, 1964), familiarization (Bateson, 1966),
and risk-aS-a-value theories (Levinger & Schneider, 1969). The diffusion of
responsibility argument contends that group participation induces risky Shift because
accountability for any negative consequences resulting from risk taking can be Shared
among group members, rather than by Single individuals (e. g., Wallach, Kogan, &

Bern, 1964). Familiarization theory argues that group discussion reduces uncertainty,

58
thereby making individuals more comfortable with risky choices (e.g. , Bateson,

1966). The risk—as—a-value explanation alleges that risk is a stronger cultural value
than caution (e. g., Levinger & Schneider, 1969). In group settings, individuals are
exposed to positions that are riskier than their own. According to the risk-aS-a-value
explanation, risky shift results because individuals attempt to be at least as willing to
accept risk in their personal decisions as other people.

Although these theories constitute an interesting area of research, they are not
‘ particularly relevant to the issue of risky decision making in a team context. They
focus on individual decision riskiness before and after group discussion. Leaders in
hierarchical decision making teams must Operate in the team context before, during,
and after decision making. ”Risky shift" seems to have been more popularized, but
the decision making literature also offers considerable evidence for conservative Shifts
(e.g., Vinokur, 1971a; 1971b). These theories are also limited because although each
can potentially explain risky shift, none of them can account for shifts toward caution.

W. The discussion in this section focuses on the small subset of
research conducted under the rubric of risky shift that (a) describes team decision
riskiness, and (b) attempts to explain both risky and cautious Shifts. This theoretical
and empirical work compares the riskiness of choices made by individuals to the
riskiness of a consensus choice (i.e., one made by mutual agreement) made by the
same individuals as a team (e.g., Hartnett & Barber, 1974; Neale, Bazerman,
Northcraft, & Alperson, 1986). Relevant theories include: subjective expected utility

(e. g., Vinokur, 1971a; 1971b), leadership/persuasion (e.g., Bumstein, 1969), group

59
polarization (Lamm & Meyers, 1978; Meyers & Lamm, 1975), and group attenuation

theories (e.g., Neale et al., 1986). Each is discussed below.

Subjective expected utility theory has been proposed to account for both risky
and cautious shifts. Vinokur (1971a, 1971b) argued that new information raised in
team discussion results in more accurate assessments of utility, which in turn, affect
the riskiness of a team’s decision. This explanation, labelled the informational
influence hypothesis, was supported by data from a series of four empirical studies
(V inokur, 1971a). Subjects in each study worked on hypothetical choice problems
independently first, then as members of four or five person groups. Results
demonstrated that groups assessed utilities more accurately in a team context and that
resulting choice shifts were appropriately more conservative or risky, depending upon
whether initial individual utilities were too risky or conservative.

While the informational influence explanation is relevant for risky decisions
made in a team context, it is limited to certain situations. Not all team discussions
result in the consideration of new information. On the contrary, more recent
empirical research suggests that discussions tend to focus on information that team
members hold in common and information that is consistent with initial preferences
(Stasser & Titus, 1985).

Originally, leadership or persuasion theory focused solely on risky shift.
Proponents of this theory claimed that high risk takers are more persuasive in group
discussions, resulting in a general shift toward risk (Collins & Guetzkow, 1964;

Marquis, 1962). Empirical evidence, however, refuted the theory in this form (see

60

Clark, 1971). The revised leadership/persuasion explanation attempts to account for
both risky and cautious Shifts. The updated theory contends that the most confident
individual exerts the most influence in a discussion (Bumstein, 1969). Therefore,
whether a risky or cautious Shift occurs depends on the orientation of the most
confident member.

Group polarization theory also accounts for both risky and cautious shifts.
According to this hypothesis, the average position or the position most commonly
held by individuals will be reinforced and become more pronounced in group
discussion (Lamm & Meyers, 1978; Meyers & Lamm, 1975). Figure 4a provides an
illustration of the polarization hypothesis. Risky Shift occurs when the average
individual stance on risk is at about Point C. It is posited that group discussion
polarizes this position, and the team makes a decision in the more extreme direction
toward Point D. A cautious Shift is evidenced when the average individual attitude
toward risk is at about Point B. Here, polarization creates a Shift toward Point A.

Recently, a phenomenon opposite to group polarization, group attenuation, has
been proposed to account for choice shift. The group attenuation notion contends that
cognitive framing is a critical variable in understanding choice shift (Hartnett &
Barber, 1974). Advocates of the group attenuation hypothesis believe that
individuals’ initial risk orientations are greatly influenced by the way the choice
problem is presented or framed (Neale et al., 1986). Group discussion is thought to
encourage the consideration of multiple frames, thereby limiting the effects of the

original frame on team decisions. In addition, discussion is expected to direct

61

 

 

Risk
Neutral
A B - C D
Rlsk I I Risk
Averse | I
Figure 4a: Illustration of group polarization effects
Risk
Neutral
A B - C D
Risk 11 I Risk
verse r l Seeking
¢———’ l——o

Flgure 4b: Illustration of group attenuation effects

62

attention away from frames and focus it on more relevant aspects of the decision
problem (Neale et al., 1986). The point of this dreary is not that the team adopts a
frame that overpowers initial individual frames. On the contrary, the theory argues
that potential framing effects are diffused and less relevant in a team context. AS a
result, teams’ decisions are expected to be attenuated relative to individuals’ more
extreme decisions that are heavily influenced by frame.

Figure 4b is useful in illustrating the expected group attenuation effects.
Relying on prospect theory research, the contention is that, under gain frames, the
average riskiness of individual decisions is represented by Point A. Under loss
frames, Point D reflects the average riskiness of individual decisions. Team decisions
are expected to Show attenuation, as attention is directed away from the decision
frame. Thus, individuals. originally at Point A Should end up at Point B as a team
after group consensus, reflecting a risky shift. Individuals originally at Point D
Should Show a cautious shift toward Point C. Note, however, that teams are still
relatively risk averse under gain frames and risk seeking under loss frames.

Empirical results relevant to the attenuation hypothesis have been mixed.
Using a negatively framed version of the CDQ, Hartnett and Barber (1974) found that
teams were even more risk seeking than individuals, supporting polarization and
refuting attenuation theory. The results of Neale et al. (1986), however, were
supportive of attenuation theory. Under positive or gain frames, teams selected
riskier alternatives than individuals. Under negative or loss frames, teams opted for

less risky alternatives than individuals. Rather than the CDQ, Neale et al. employed

63
adaptations of Kahneman and Tversky’s (1979) prospect theory problems. Referring

back to Figures 4a and 4b may help reconcile these findings and the differential
predictions of polarization and attenuation theories.

The direction of the shift in risk when team and individual decisions are
compared, appears to be constrained by the initial individual average. For example,
if the average is represented by point A there is not much room to Show a Shift
toward greater caution. A shift toward increased riskiness is much more likely.
Similarly, if the individual average is represented by Point D, there is less chance to
Show a risky Shift than a cautious Shift. Results under these conditions would appear
to support attenuation theory.

Now consider Points B and C as indicators of the average risk of individual
decisions. If individuals are initially risk averse at Point B, it appears as though a
team decision could Shift risk in either direction. However, assuming that prospect
theory does account for choice shift as the proponents of attenuation theory claim
(Bazerman, 1984; Neale et al., 1986), a Shift toward point A is much more likely
than a shift toward Points C and D. The rationale is as follows. The average of
individual decisions should be risk averse under gain frames. To remain consistent
with the gain frame, the team’s decision can only move to Point A. There is no
empirical or theoretical evidence to suggest that teams’ reactions to frames should be
the opposite of individual reactions. The same logic would suggest a greater
likelihood of a choice Shift from Point C to Point D, compared to the likelihood of a

shift from Point C toward Points A and B.

64

It appears as though the group polarization and group attenuation theories use
similar logic to arrive at contradictory conclusions. In actuality the theories are not
inconsistent, they simply describe two different potential sets of circumstances. The
within groups nature of the risky shift studies comparing individuals to teams makes
drawing conclusions about the relative riskiness of team decisions compared to
individual decisions difficult. For present purposes, the primary usefulness of both
theories comes from demonstrating that the riskiness of team decisions does not differ
radically from the riskiness of individual decisions. On the basis of the evidence cited
above, it appears that, regardless of whether team decisions become polarized or
attenuated, they remain consistent with prospect theory framing predictions, just as
individual decisions do.

mm. The majority of risky Shift research is concerned with individual
decisions and has little relevance for risky decisions made in a team context (e.g.,
Bateson, 1966; Levinger & Schneider, 1969; Wallach et a1. , 1964). The applicability
of the work examining risk as it varies across individual and team decisions to
hierarchical teams with distributed expertise must be considered cautiously. Risky
Shift research has focused exclusively on consensus decisions made by teams. Even
though the contribution is still minimal, the group polarization and group attenuation
theories have the greatest potential relevance to decision making in hierarchical teams
with distributed expertise.

The polarization and attenuation theories both suggest that decision frames

operate in both individual and team decision making contexts. Since an individual

65

(i.e. , the leader) makes the decisions in a hierarchical team, one might expect
decision frames to have a similar effect. The risky shift literature suggests that
prospect theory framing effects can occur in collectives. Empirical research
attempting a more direct test of prospect theory in the team context is discussed in the
section that follows.

Win—item

An argument was made above for the existence of framing effects in team
decision making contexts. Using the polarization argument, some have even
contended that the effects of framing are likely to be more pronounced in teams (e.g. ,
Whyte 1989). The problems inherent to a within groups comparison of the riskiness
of individual and team decisions were previously discussed. A few studies have
attempted to directly assess framing effects on risky decision making in team contexts.
Those studies are described below.

Schurr ( 1987) purported to study the effects of framing on risky team decisions
involving a bargaining task. However, in my opinion, Schurr failed to accomplish
this objective because he did not use teams in his research. Each participant played
the role of a commodity broker. Individual ”brokers" assigned to different teams
negotiated with one another for a specific quality level for each commodity. The goal
was to maximize net profit. The winning team was determined by totaling the net
profit associated with each individual transaction.

The study’s results were supportive of prospect theory predictions for both

gain and loss frames. However, in my opinion, the results cannot be interpreted at

66

the team level. Schurr’s intention was to study risky team decision making, yet
individuals worked on the bargaining task independently. Subjects were simply told
that they belonged to one team or the other and that individual bargaining profits
would be combined to measure team performance. The definition of a team utilized
here requires that individual team members interact. The subjects in Schurr’s study
did not interact with assigned teammates at all. Consequently, Schurr’s results do not
speak to risky decision making in a team context. They do, however, demonstrate
that the effects of framing on complex decision making tasks are consistent with
results obtained using simple choice tasks.

Schurr’s study included a unique conceptualization and operationalization of
risk and a complex decision task. Tversky and Kahneman’s (1981) work and most of
the empirical research on prospect theory has focused on whether individuals select a
certain or a risky alternative. The bargaining task allowed Schurr to explore degrees
of risk taking on a continuum of risky outcomes. In addition, decision makers were
actively engaged in a decision task substantially more complex than the typical
prospect theory choice problem. Thus, Shurr’s study demonstrated the effects of
framing on risky decision making for individuals working on a task more complex
and engaging than the traditional binary choice task. However, this research does not
fill the need for empirical work related to the team decision making context.

McGuire, Kiesler, and Siegel (1987) studied the effects of framing on risky
decision making in face-to—face and computer mediated groups using a complex

budgeting task. In addition to comparing risky decision making using the two modes

67

of communication, this study sought to test explicitly the notion that individual level
framing effects become more pronounced in a team context. The budgeting task
consisted of four hypothetical scenarios (two gain and two loss) which each offered
two investment alternatives. Each individual was asked to first indicate a preference
between alternatives privately, before attempting to reach a consensus decision with
team members.

Results related to the first phase, the individual decisions, did not consistently
conform to prospect theory predictions. Therefore, it was not possible to determine if
individual level framing effects were exacerbated at the team level. McGuire et a1. ’8
explanation for the lack of conformity to prospect theory predictions was that some
subjects misunderstood the decision task. Another possibility is diffusion of
treatments. Research at the individual level has Shown that the effects of initial frame
persist, even when a new frame is presented (e.g., Levin, Johnson, & Davis, 1987;
Loke, 1989). Nonetheless, decision behavior in face-to-face teams did conform to
prospect theory predictions. Framing, however, had no effect in the computer
mediated teams. Under both modes of communication, the best predictor of a team’s
final decision was the judgment or position first advocated.

This study’s results are a bit difficult to interpret. The differential findings
related to framing in the computer mediated and face-to-face conditions could not be
adequately explained. Since teams were required to reach consensus, different
theories derived from the consensus literature were offered to explain the "first

position advanced" results (e.g., the individual with the most influence states a

68

position first, individuals conform to the voiced alternative). However, none of the
explanations could be verified. Essentially, this study demonstrated the existence of
framing effects in a team context. It also demonstrated that the effects are not
consistent and must be subject to certain boundary conditions and moderating
influences. Unfortunately, this particular piece of research offers few clues as to
what the limiting conditions and moderators may be. It Simply identifies the need for
further research using computer mediated technologies.

One final study examined the effects of actual loss on risky decision behavior
in a team context. Kameda and Davis (1990) examined the effects of individual loss
on subsequent decision making as part of a team. Using a gambling task which
offered actual monetary outcomes, they found that when placed in a team, individuals
who had experienced a loss wanted to make riskier choices than individuals who had
not experienced loss. However, these individuals were not able to influence nonloss
members and were overridden by the majority. In this instance, team membership
prevented risky decision making largely because of majority rule. This study Shows
that loss in an individual’s performance history does not necessarily determine
subsequent team decisions made on the basis of majority rule. The implication,
however, is that if team members have similar performance histories the impact on
subsequent risky decisions may be great. In the discussion of individual level risky
decision making presented earlier, performance history was expected to have a direct

impact on perceived control.

69
The research reviewed above dealt with framing effects for teams in which

decisions were made by consensus (i.e., McGuire et al., 1987) and majority nrle
(i.e., Kameda & Davis, 1990). Recall, however, that the present research will focus
on hierarchical teams with distributed expertise in which the team leader makes the
team decision. Therefore, a key issue is predicting the impact that the introduction of
”hierarchy" is likely to have on observed framing effects. In a hierarchical team, the
most relevant frame is likely to be that of the decision maker, the team leader. The
research of Kameda and Davis (1990) suggests that the leader’s performance history
will have considerable influence on the riskiness of the decisions made. The team
decisions in their study were consistent with majority rule. The leader of a
hierarchical team, however, is not required to accommodate the majority and can
make decisions on the basis of his or her own perceptions of the frame.

The sparse research conducted to date has detected the framing effects
predicted by prospect theory in team contexts, but not consistently. There is an
obvious need for team research utilizing complex and engaging decision tasks (e.g. ,
Schurr, 1987). There is also some indication that computer mediated technologies
Should be explored further (e. g., McGuire et al., 1987). These needs will be
addressed as a model for risky decision making in hierarchical teams is developed.
First, however, one additional literature will be considered for application to risky
decision making in hierarchical teams. Since team leaders bear decision making
responsibility in hierarchical teams, selected literature on leader decision making will

be discussed in the following section.

70
Leader Decision Making

The Vroom-Yetton contingency model of leader behavior is probably the most
widely known model of leader decision making in the organizational literature. It is
also indirectly applicable to present research purposes. Developed by Vroom and
Yetton (1973), the model specifies the type of decision making strategy that should be
used by a leader, based on an assessment of the situational demands. More
specifically, the model outlines seven rules to be used in choosing among five
decision making processes. The rules are broadly divided between two categories,
those that protect the quality of the decision and those that protect the acceptance of
the decision (see Table 1). The five decision making processes enumerated in Table 2
range from making a completely autocratic decision (i.e. , the leader decides alone on
the basis of information available at a given point) to making a totally participative
decision (i.e. , subordinates generate and/or consider alternatives and the leader
accepts any agreed upon decision).

The Vroom-Yetton model provides a leader with a useful tool that may be
used in determining the best method of making a decision under a given set of
circumstances. Thus, in one sense, the model is more prescriptive than descriptive.
The model developed here is meant to be a description of risky decision making in a
hierarchical team context. Even though there is not a perfect marriage between the
goals of the Vroom-Yetton model and those of the present model, the rules offered by

the Vroom-Yetton model can serve as catalysts for thinking about the different

71

Table 1

8112 Eng}.

1.

The Leader Information Rule: If the quality of the decision is important and the
leader does not possess enough information or expertise to solve the problem by
himself, then AI is eliminated.

The Goal Congruence Rule: If the quality of the decision is important and
subordinates are not likely to pursue the organization goals in their efforts to
solve this problem, then GII is eliminated.

The Unstructured Problem Rule: In decisions in which the quality of the decision
is important, if the leader lacks the necessary information or expertise to solve
the problem by himself, and if the problem is unstructured, the method of solving
the problem should provide for interaction among subordinates likely to possess
relevant information. Thus, A1, A11, and C1 are eliminated.

f D ' ° n
The Acceptance Rule: If the acceptance of the decision by subordinates is critical
to effective implementation and if it is not certain that an autocratic decision will
be accepted, AI and A11 are eliminated from the feasible set.

The Conflict Rule: If the acceptance of the decision is critical, an autocratic
decision is not certain to be accepted and disagreement among subordinates in
methods of attaining the organizational goal is likely, the methods used in
problem solving should enable those in disagreement to resolve their differences
with full knowledge of the problem. Accordingly, under these conditions, A1,
A11, and CI, which permit no interaction, are eliminated.

The Fairness Rule: If the quality of the decision is unimportant but acceptance
is critical and not certain to result from an autocratic decision, it is important that
the decision process used generate the needed acceptance. The decision process
used should permit the subordinates to interact with one another and negotiate
over the fair method of resolving any differences with full responsibility on them
for determining what is fair and equitable. Under these circumstances, A1, A11,
CI, and CH are eliminated.

The Acceptance Priority Rule: If acceptance is critical, not certain to result from
an autocratic decision, and if subordinates are motivated to pursue the
organizational goals represented in the problem, then methods offering equal
partnership in the decision-making process can provide greater acceptance without
risking decision quality. Thus, AI, All, C1, and C11 are eliminated.

Adapted from Vroom and Jago (1978).

72

Table 2

ENE

A1

All

C1

C11

GII

You solve the problem or make the decision yourself using the information
available at present.

You obtain necessary information from subordinates, then decide on a solution
to the problem yourself. You may or may not tell subordinates the purpose of
your questions or give them any information about the problem. The input
provided by them is clearly in response to your request for Specific information.
They do not play a role in the definition of the problem or in generating or
evaluating alternative solutions.

You Share the problem with the relevant subordinates individually, getting their
ideas and suggestions without bringing them together as a group. Then you make
the decision. This decision may or may not reflect your subordinates’ influence.

You share the problem with your subordinates in a group meeting. Your obtain
their ideas and suggestions. Then, you make the decision, which may or may not
reflect your subordinates’ influence.

You share the problem with your subordinates as a group. Together you generate
and evaluate alternatives and attempt to reach agreement (consensus) on a
solution. Your role is much like that of a chairman. Your can provide the group
with information or ideas that you have but you do not try to ”press” them to
adopt "your" solution and are willing to accept and implement any solution that
has the support of the entire group.

Adapted from Vroom and Jago (1978).

73

methods a leader of a hierarchical team may use to make the team’s decisions and the
conditions under which certain methods are more likely to be utilized.

A list of different leader decision strategies proposed by the Vroom-Yetton
model are shown in Table 2. Although the individual strategies as Stated are not
directly relevant to leaders in hierarchical teams, taken as a whole they do provide
some indication of the range of potential methods available to a leader making a team
decision. For instance, a leader in a hierarchical team with distributed expertise may
choose to obtain information from subordinates and make a decision on the basis of
that information alone. At the other end of the continuum, the leader may rely solely
on the recommendations of subordinate team members in making a decision. Again,
the taxonomy of decision processes presented in Table 2 is not necessarily applicable
to hierarchical teams as specifically stated, but it can serve as a heuristic for thinking
about the way decisions are made.

The list of rules for determining decision strategy provided in Table 1, may be
helpful in considering differential decision making strategies under gain and loss
frames, to the extent that they describe characteristics of the two conditions. For
instance, the Acceptance Priority Rule describes conditions where support for a
decision is necessary and unlikely, conditions more likely to exist under a loss frame
than a gain frame. When there is only potential to lose the team’s existing outcomes
(i.e. , loss frames) acceptance may be more important as compared to a Situation in

which there is only potential to gain additional outcomes (i.e. , gain frames). Under

74

these circumstances the Vroom-Yetton model recommends a leader decision strategy
that relies heavily on subordinates. In the present context, that could mean relying
more on subordinate judgements than raw informational cues. The Fairness Rule,
which suggests the same type of decision strategy as the Acceptance Priority Rule,
also seems likely to be invoked under a loss frame.

The Vroom-Yetton model has the potential to provide insight into how
decisions are made in a team context. The empirical literature based on the model is
also encouraging. First, research has shown that the five decision processes outlined
by the model do result in predicted effects on decision quality and decision acceptance
(Field, 1982; Vroom & Jago, 1978). In addition, leaders report that their own
decision making practices conform to the decision making processes recommended by
the model (Field & House, 1990). Finally, the rules for choosing a decision strategy
are perceived as appropriate and valid, at least by leaders (Heilman, Homstein, Cage,
& Herschlag, 1984).

In the section that follows, a model of risky decision making in the team
context is developed. The model draws on the literatures that consider risky decision
making in individual and team contexts. The model also utilizes the leader decision
making literature. The specific contributions of each literature will become more

apparent as hypotheses related to decision making in the team context are derived.

RISKY DECISION MAKING IN HIERARCHICAL TEAMS

A primary goal of the present research is to Study cognitive framing effects
and their boundary conditions in the team decision making context. In the preceding
section, literature with potential relevance to comprehending risky decision making in
hierarchical teams was explored. The review of the risky shift and ”team” level
prospect theory literatures demonstrated a general paucity of research focused on
risky decision making in the team context. No theoretical attention has been paid
specifically to risky decision making in hierarchical teams, even though this type of
team represents the next logical Step in a progression moving from the individual to

the team level. The Vroom-Yetton model of leader decision making was reviewed,

demonstrating some indirect linkages to the present research.

 

A preliminary definition of hierarchical teams with distributed expertise was
provided previously. However, a more detailed description is required to begin
developing a model of risky decision making for leaders in hierarchical teams. In the
sections that follow, both the characteristics of hierarchical teams in general and the
more specific characteristics of the hierarchical teams of interest here are discussed.

W. Hierarchical decision making teams have three
primary characteristics: status differences (i.e., hierarchy), distributed expertise, and
a communication structure (Ilgen et a1. , in press). In a hierarchical team, members
do not have equal status. In the Simplest case, there are two levels to the hierarchy,

leader and subordinates. The present research will focus on teams where status

75

76

among subordinates is equal, but this is not a definitional characteristic of all
hierarchical teams. The leader has ultimate decision making responsibility. All
members of the team (i.e. , the leader and each subordinate) are subject to the
consequences of the leader’s decision. Thus, unlike an individual decision maker, the
team leader is accountable to others. The subordinates, all of whom have lower
status than the leader, may contribute to the decision by offering the leader
recommendations. That is, subordinates may suggest that the leader make a certain
decision. Whether or not subordinate judgments are taken into account in making a
final decision, however, is at the discretion of the leader. Brehmer and Hagafors
(1986) described this type of hierarchical decision making structure as ”staff decision
making. "

AS discussed in a previous section, teams structured on the basis of distributed
expertise have clearly defined and differentiated roles. Each team member is
considered an expert in certain aspects of the decision task and is responsible for
attending to those aspects. As a result, no individual team member, including the
leader, is likely to be knowledgeable in all aspects of the decision task. To a large
extent, expertise is determined by the allocation of information across team members.
Teams in which each member has access to all relevant information have redundant
structures. They are not structured according to distributed expertise. In teams with
distributed expertise, members have direct access to some information relevant to
their areas of expertise, but they must interact and communicate with other members

to have complete information. Thus, decision making in a team context has process

77

costs (i.e. , required interactions) that are not necessary for individual decision
makers.

The final characteristic, communication structure, is very important. A team’s
communication structure defines which members may communicate directly and which
are linked only indirectly. In many cases, each member of a team can communicate
directly with every other member of the team. However, it is also possible for
communication to be more limited. For example, subordinates may only be able to
communicate with one another by going through the leader.

SW. The decision making task is described in detail in the
method section, however, to grasp the implications of this particular information
processing simulation for the team model, some of the features must be mentioned
here. In the present research, each experimental condition is comprised of four-
person teams (i.e. , a leader and three subordinates) that make decisions under a
hierarchical structure with distributed expertise. Subordinates have direct access to an
equal number of informational cues, but must communicate with one another in order
to have complete information about their areas of expertise. The leader has direct
access to some, but not all informational cues. The leader, however, knows which
team members have direct access to which informational cues. The communication
Structure will be such that each team member can communicate with all others. In
each case, subordinates will be able to provide recommendations to the leader.

Responsibility for the team’s decision, however, is the leader’s.

78
Critical Features of the Team Context

Great care must be taken in applying the individual model of risky decision
making in order to understand risky decision making in the team context. The
previous chapter assessed several literatures to determine their relevance for risky
decision making in the team context. The definitional characteristics of teams in
general and hierarchical teams with distributed expertise in particular have also been
discussed. The remaining objective is to distinguish those critical features of the team
context that are likely to have implications for the riskiness of decisions made in
teams.

Since the present research is expressly interested in the context created by
hierarchical teams with distributed expertise, the team context issues that may
influence the team leader’s decisions are the focus. Decision cues, the decision
making process, and decision outcomes in a team context are substantially different
from those in an individual decision making context. Each of these aspects is
considered below.
misign gigs

Individual decision makers process informational cues in order to reach a
decision. Team leaders also process informational cues, but in addition they may
often have recommendations from other team members to consider. In essence,
recommendation cues are suggestions or judgments offered to influence the team

leader’s decision.

79

. Recommendation cues are qualitatively different from informational cues.
They are evaluative rather than objective. The recommendations can vary in their
accuracy. Taken collectively, the recommendations offered by subordinate team
members can vary in their consistency (i.e. , individual subordinates may make
different recommendations). One may think of recommendation cues as ”active” in
that the cue providers (i.e. , subordinate team members) are able to respond to the
team leader. All of these qualities make using recommendation cues different than
using informational cues to make a decision.

For instance, a leader may assess the quality of recommendation cues before
using them. The recommendation one can be accurate or inaccurate, as can the
leader’s evaluation of it. If a team leader receives conflicting recommendations,
using the cues becomes even more problematic. There may be repercussions resulting
from using one subordinate’s recommendation instead of another’s. The way a leader
handles recommendations provided for one decision may affect the type of
recommendations offered for future decisions.

Specific hypotheses regarding recommendation cues are provided in a
subsequent section. The primary point here is that recommendation cues are unique
to the team decision making context. Such cues are not relevant for individual
decision makers acting alone.

' i n 'n
The notion of distributed expertise was considered previously. Hierarchical

teams with distributed expertise are structured in such a way that the decision maker

80

in that context is, in a sense, "dependent” on subordinate members. By definition,
distributed expertise means that no Single team member, including the leader, is
expert in all areas relevant to the decision task. Instead, the expertise is disbursed
among team members. A team leader, then, must rely on subordinate members to
provide and interpret information relevant to their particular areas of expertise. An
individual decision maker is not dependent upon others in this way. This is another
way in which the team context of interest here differs from the individual decision
making context.
i i m

The leader of a hierarchical team with distributed expertise may be dependent
upon subordinate members for their expertise, but subordinate members are also, in a
sense, dependent on the team leader. The team leader has ultimate decision making
authority. The leader’s decision and all its ramifications stands for the entire team.
Thus, any outcomes associated with the leader’s decision apply not only to the leader,
but to all members of the team.

For an individual decision maker, a decision has only individual consequences.
In the team context, a decision has consequences for others. These are individuals
upon whom the leader must rely, individuals who may respond and react to the
leader’s decision and its outcomes. Thus, the implications of a decision are different
for a team leader. There are potential costs and benefits associated with the decision

that are not relevant for an individual decision maker.

81

These are the type of context issues that may have implications for the
decision making of a leader in a hierarchical team. These factors distinguish the team
decision making context from the individual decision making context. This is not to
say that decision making in a team context is radically different from individual
decision making, in all respects. AS the team context model is developed in the
following section, the ways in which individual and leader decision making are
similar and different will be elaborated more fully.

The Team Context Model of Risky Decision Making

The central focus of the model of individual risky decision making developed
previously was the hypothesized moderating effect of perceived control on the
relationship between cognitive decision frames and decision riskiness. Perceived
control is also proposed as moderator of the same relationship in the model developed
for the risky decision making of team leaders. However, in elaborating and
distinguishing the team leader model, it is also important to consider the unique
factors that may moderate the relationship between decision frames and decision
riskiness only in the team context. Two such moderators, need for affiliation and
cohesiveness, are advanced below following the discussion of perceived control for
team leaders.

W

W. One reason for the consideration of individual risky

decision making was to provide a basis for the development of a model relevant to

risky decision making in hierarchical teams. The core of the individual model is

82
composed of two key linkages. The first is the relationship described by prospect

theory, the linkage between cognitive decision frame and decision riskiness. The
second is the moderating effect of perceived control on the relationship between
decision frame and decision riskiness.

This core serves as a starting point for the team context model of risky
decision making. Framing effects have been demonstrated at the team level (e.g. ,
McGuire et al., 1987) and are expected to persist for decisions made in a hierarchical
team context. To reiterate, team leaders are expected to be risk averse under gain
frames and risk seeking under loss frames. There is some suggestion in the literature
that framing effects may be more pronounced for teams than for individuals (e.g. ,
Whyte, 1989). This is largely an empirical question that can be tested.

There is also some indication that the leader’s perceived control may have a
Significant impact on the riskiness of team decisions (e. g., Bumstein, 1969). The
predicted moderator effect can be tested empirically, but there is less chance of
detecting it in the team context. A team leader’s Situation is substantially different
from that of an individual decision maker. The leader’s perceptions of control are
likely to be affected by the team context. The leader has ultimate decision making
responsibility and the outcomes of those decisions affect others. Subordinates are
likely to react to the outcomes they receive. They may have advice for the leader, in
addition to the capability to express their satisfaction or dissatisfaction with the
decisions a leader makes. Essentially, the team context may create ”noise” not

present at the individual level. Therefore, even given equal levels of power, it is

83

likely to be more difficult to detect a moderator effect in the team context. The
nature of perceived control’s moderator effect in the individual model (see Figure 2)
is also hypothesized to apply in the team context model.

Won. Need for affiliation is an individual difference variable
that was not included in the individual model of risky decision making. The need for
affiliation, along with the need for achievement and the need for power, comprise the
trichotomy of needs that drive human behavior according to McClelland’s theory of
human motivation (e. g., McClelland, 1961; 1975; McClelland & Winter, 1969). The
need for affiliation refers to an individual’s desire to be liked by others. This trait
seems to be particularly relevant for leader’s making risky decisions in a team
context.

Previous research has Shown that leaders (e. g. , effective managers) tend to
have a particular motive pattern. Of most importance for present purposes, is the
finding that effective leaders tend to have a low need for affiliation coupled with a
high nwd for achievement (McClelland & Boyatzis, 1982). Apparently a lower need
for affiliation allows a leader to make and execute important and sometimes unpopular
decisions without being hampered by concern about being liked by others. A leader
in a hierarchical team faced with a risky decision is responsible for the outcomes of
subordinate team members. The decisions the leader makes affect those outcomes. A
leader who has a higher nwd for affiliation is more likely than a leader lower in the

need for affiliation to try and make decisions that satisfy subordinates.

34

According to the threat literature, people perceive potential gains as
opportunities and potential losses as threats (Dutton & Jackson, 1987 ; Jackson &
Dutton, 1988). A leader with a high need for affiliation is expected to take advantage
of opportunities by securing certain gains for team members. Leaders higher in the
nwd for affiliation are also expected to thwart threats to the team by making every
effort to avoid the loss of existing outcomes. Stated in terms of risky decision
making under cognitive frames, leaders with a higher need for affiliation are expected
to be risk averse under gain frames and risk seeking under loss frames. On the other
hand, Since leaders lower in the need for affiliation are supposed to be more effective
leaders (McClelland & Boyatzis, 1982) they are more likely to focus on maximizing
outcomes regardless of decision frame. Specifically, leaders lower in the need for
affiliation are predicted to make more risk seeking decisions (i.e. , outcome
maximizing decisions) regardless of the cognitive frame (i.e., loss or gain). The
nature of the moderating relationship is shown in Figure 5.

Cghgsimgss. Cohesiveness is another construct that is unique to risky
decision making in the team context. Since cohesiveness is a team characteristic, it
was not relevant to the individual level model. The problems associated with defining
cohesiveness are well documented (Mudrack, 1989). Like many constructs in the
psychological literature, the term cohesiveness has been defined in so many ways and
used with such generality that it has become virtually meaningless in many contexts.
Nonetheless, the problem cannot be solved by coining a new term in an attempt to

avoid the ”baggage" associated with the label, cohesiveness. Instead, the original and

85

, Low Need for Affiliation
High

 

RISK

High Need for Affiliation
Low

 

 

Gain Loss
DECISION FRAME

Figure 5: Hypothesized moderating effect of

need for affiliation

86

most relevant definition of cohesiveness will be employed. Cohesiveness was
originally and is still most frequently defined as attraction to the group, or in this
case, attraction to the team (Libo, 1953; Pepitone & Kleiner, 1957 ; Van Bergen &
Koekebakker, 1959).

Empirical research has shown that cohesiveness is related to a number of
important outcome variables. For example, several laboratory studies have found a
positive relationship between cohesiveness and performance (e.g. , Bakeman &
Helmreich, 1975; Bird, 1977; Dorfman & Stephan, 1984; Hoogstraten & Vorst,
1978; Zaccaro & McCoy, 1988), a relationship also supported in a recent meta-
analysis (Evans & Dion, 1991). Cohesiveness is also positively related to more
affective outcomes, such as satisfaction (Dailey, 1978; Dobbins & Zaccaro, 1986) and
perceived social support (Griffith, 1989).

In the empirical study most directly relevant to the present research purposes,
Yinon & Bizman (1974) examined the relationship between cohesiveness and risk
talcing. Their study employed three-member teams in which one individual was
responsible for the making the team’s decision. They found that decision makers in
more cohesive teams made less risky decisions than decision makers in less cohesive
teams. They argued that decision makers in highly cohesive teams had more to lose
socially by subjecting their teams to failure. Thus, they tended to be more risk
averse. On the other hand, they contended that a decision maker in a less cohesive

team could only improve the team’s circumstances through achieving successes (i.e. ,

87

maximizing outcomes). Therefore these decision makers behaved in a more risk
seeking manner.

Based on Yinon and Bizman ’s ( 1974) results, the present research will
examine the interaction between cohesiveness and cognitive decision frames in
predicting decision riskiness. Leaders of highly cohesive teams should be motivated
to avoid failure. Therefore, under gain frames they are likely to be risk averse.
Avoiding failure under loss frames may mean that they cannot accept the certain loss.
Thus, leaders of highly cohesive teams are likely to be risk seeking under loss
frames. Leaders of less cohesive teams may be more concerned with maximizing
outcomes and less concerned with maintaining attraction bonds. Thus, leaders of less
cohesive teams are predicted to make more consistently risk seeking decisions
regardless of decision frame. The moderating effect of cohesiveness is depicted in
Figure 6.

The inclusion of cohesiveness in the team context model raises a measurement
issue that was not pertinent until this time. There are several ways to measure the
team characteristic, cohesiveness. One way would be to take a measure of
cohesiveness from each team member and aggregate (i.e. , average) those ratings.
Another way would be to decide which member is most representative of the team
and use his or her rating of cohesiveness in the analyses. One other method, the one
used in this study, is to consider the hypothesis and choose the measure that seems
most relevant. In this case, cohesiveness is expected to impact on the relationship

between frame and the riskiness of decisions. Since the decisions are made by the

88

Low Cohesiveness
High

 

RISK

  

High Cohesiveness

Low

 

 

Gain Loss
DECISION FRAME

Figure 6: Hypothesized moderating effect of cohesiveness

89

team leader, the leader’s rating of cohesiveness is used. Cohesiveness is
conceptualized as attraction to the team. Given the hypothesis, the leader’s attraction
to the team seems most germane.
W

Figure 7 shows the direct linkage from cognitive decision frame to decision
riskiness and the three proposed moderators of that relationship: perceived control,
need for affiliation, and cohesiveness. Recall that under gain frames leaders with
higher perceived control are expected to make riskier decisions than those with lower
perceived control. leaders are expected to be consistently risk seeking under loss
frames, regardless of perceived control. Leaders with a lower need for affiliation are
predicted to make more risk seeking decisions (i.e. , outcome maximizing decisions)
regardless of the cognitive frame (i.e. , loss or gain). The decisions of those with a
higher need for affiliation are expected to conform to the predictions of prospect
theory (i.e. , risk seeking under loss flames and risk averse under gain frames). The
decisions of leaders in more cohesive teams are also expected to conform to prospect
theory predictions. Leaders of less cohesive teams are predicted to make more
consistently risk seeking decisions under both gain and loss frames. Figure 7 captures
the basic model of team leaders’ risky decision making. To clarify measurement
considerations, a superscript is attached to each variable indicating whether that

variable is measured at the decision (D) or the individual (I) (i.e. , team leader) level.

 

PERCEIVED CONTROLD
NEED FOR AFFILIATIONI
TEAM COHESIVENESSI

 

 

 

 

 

 

 

DECISION FRAME I _._. DECISION RISKD

 

 

 

 

 

Figure 7: Primary model of risky decision making
in a team context

91
Predictors of Leader Perceived Control

At the individual level, predictors of perceived control were considered in the
model of risky decision making. The team context model is much too preliminary for
anything more than a speculative consideration of such factors. However, to maintain
symmetry between the models, the factors that may influence a team leader’s
perceptions of control will be considered from an exploratory perspective.

1i ' n f h In ivi M 1

Once again, the individual model will be used as a starting point. Four factors
were considered at the individual level: locus of control, task knowledge,
performance history, and one ambiguity. These same factors are also considered for
teams, although their treatment is somewhat more problematic. As previously
mentioned, there is likely to be less variance in perceptions of control of team
leaders. Thus, there will be less variance in the criterion when perceived control is
the dependent variable. The hypothesized predictors were fairly straightforward at the
individual level, but they will be more complex as team context is considered.

W. Locus Of control refers to whether events and outcomes are
more likely to be viewed as contingent upon personal behavior or seen as caused by
external forces (Rotter, 1966). Those toward the internal end of the scale tend to
believe the former, while those closer to the external end are prone to maintain the
latter. At the individual level, decision makers with a more internal locus of control
were expected to perceive greater control over decision tasks than extemals. Internals

were expected to view themselves as having a greater probability of making a correct

92

decision. The position taken here is that, in a hierarchical team where the ultimate
authority for the team’s decision rests with the leader, it is the leader’s locus of
control that is important to the team’s decision.

As at the individual level, leaders with a more internal locus of control are
expected to have greater perceived control than those with a more external locus of
control. However, the magnitude of locus of control’s influence is not likely to be as
strong in the team context as at the individual level. While likely to be somewhat
influenced by locus of control, a personal characteristic, the perceived control of a
team leader is probably more susceptible to situational aspects of the task and team
structure. Even though the decision processes are somewhat similar for individuals
and team leaders, the leader is continuously confronted with the team context and all
its implications. The leader of a hierarchical team with distributed expertise must
deal with the team in getting decision relevant information, in making an actual
choice, and in receiving the resulting outcomes. The team is the most salient issue.

W. Task knowledge refers to the decision maker’s level of
expertise on the decision task. For an information processing task, expertise or task
knowledge is reflected in the extent to which the meanings of informational cues are
understood, the extent to which relationships among cues are understood, and the
extent to which an individual understands how the above information should be
combined and utilized to make a decision. At the individual level, greater task
knowledge was expected to be associated with greater perceived control or increased

belief in a high probability for making a correct decision.

93
While a team leader’s task knowledge Should also be positively related to

perceived control, the strength of this relationship is likely to be dependent upon the
distribution of information. At least initially, a leader in a team with distributed
expertise, knows fewer informational cues and fewer interactions among cues than an
individual decision maker who has direct access to all cues. The leader is responsible
for only a specific portion Of the decision task. Additional information can potentially
be acquired from other team members, but the information cannot be guaranteed (i.e.,
by direct access). (More will be said about information sharing later.) Even if
obtained, the use of the information is likely to be less effective because it is outside
the leader’s specific area of expertise. Therefore, the relationship between task
knowledge and perceptions of control is likely to be weaker for leaders of distributed
expertise teams than for individuals.

W. Performance history refers to the pattern of results or
outcomes resulting from previous decision making on a particular task. For
individuals, positive feedback or a positive performance history was expected to be
Strongly related to increased perceptions of control. Here, the relationship is expected
to be just as strong.

A team’s performance history provides more than a summary of correct
choices and acquired outcomes. A positive performance history informs the leader
that the Strategy being used to make decisions is effective. This is critical information
because the leader can make the decisions using a variety of criteria and programs,

some of which are outlined in the Vroom-Yetton model of leader behavior (Vroom &

94

Yetton, 1973). For instance, a leader may simply make decisions based on the
average of subordinates’ judgments. The leader may base decisions solely on the
informational cues, ignoring judgments completely. The leader may also use some
weighted combination of judgments and cues to make a decision.

Clearly, just deciding how best to make a decision is a complicated process for
team leaders. Thus, positive information indicating that whatever strategy being used
is effective is likely to go a long way to enhance perceived control. In contrast,
individual decision makers only have informational cues at their disposal. Therefore,
deciding how to make a decision is a less critical issue for individuals.

. Cue ambiguity refers to

 

the extent to which informational cues are clear or unclear in terms of their
implications for making a correct decision. A cue can be ambiguous because its
individual meaning is unclear or because it conflicts with other informational cues.
At the individual level, decision makers faced with ambiguous cues were expected to
perceive less control than decision makers confronted with unambiguous informational
cues. A similar relationship is predicted to hold for distributed expertise teams, but it
is expected to be more pronounced and more complicated as compared to the
individual level.

The increased complexity results from the fact that, in the team context, two
different types of cues may be more or less ambiguous. Recall that team leaders may
utilize informational cues and judgment cues in making a decision for the team.

Informational cue ambiguity was described previously. Judgment cues are ambiguous

95

to the extent that they do not converge. For example, consider an investment
scenario in which the leader must decide whether to invest $100, $500, or $1 ,000 for
the team. Assume that there are three subordinates and that each one advocates a
different investment amount. Under these circumstances, the team leader is
confronted with ambiguity. That ambiguity can be exacerbated to the extent that
informational cues are also ambiguous.

Contrast the scenario described above with one in which all three subordinates
recommend investing $1 ,000. Convergent judgments are expected to operate like a
preponderance of the evidence. Under these circumstances, there is little ambiguity,
especially if the informational cues are also straightforward. The team leader Should
perceive much greater control, feeling that the probability of making a correct
decision is quite high.

Ambiguous cues of both types may also erode perceptions of control in another
way. Curley, Yates, and Abrams (1986) contended that ambiguity is aversive because
it makes behavior difficult to justify to others. In the team context, the relevant
others are subordinate team members. A team leader is likely to feel less confident
and less control in making a decision on the basis of ambiguous cues because the

justification for the decision will be less evident than for a decision based on

unambiguous cues.

 

Having considered the predictors of perceived control suggested by the

individual model, potential predictors unique to the team context may now be

96

explored. The decision making process of a leader in a team context is qualitatively
different from the process of an individual acting alone. Certain team phenomenon
impinging on the leader may affect perceptions of control. The discussion that
follows addresses two factors unique to decision making in a team context that may
influence perceptions of control: centralization and coordination.

Wm. Centralization is a phenomenon unique to the team decision
making context. Here, centralization refers to the extent to which communication
within a team is concentrated around the team leader. In certain hierarchical teams,
communication linkages may be restricted such that all team members cannot directly
communicate with one another (see Ilgen et al. in press for a complete discussion).
For present purposes, however, recall that unrestricted communication paths (i.e. ,
each team member can communicate with all other team members) will be used.

Even when communication is unrestricted, it may still follow an identifiable
pattern. To the extent that the majority of communication is directed toward the team
leader, communication is centralized. For distributed expertise teams, the pattern of
communication may be a critical determinant of the leader’s perceived control.

In the section on task knowledge, a leader’s task knowledge was described as a
partial function of the extent to which other team members share their unique
informational cues and expertise with the leader. The extent to which this type of
sharing exists is represented by the centralization of communication around the leader.
Thus, task knowledge and centralization are partially redundant. Only partially,

because the leader’s task knowledge also depends upon the leader’s understanding of

97
the shared information. Like task knowledge, centralization ought to have a direct

impact on perceived control. Under high centralization, perceptions of control ought
to be greater than under conditions of low centralization.

Coordinatjgn. By definition, a team with distributed expertise must share
information if team members, in particular the team leader, are to have access to all
informational cues. In this context, coordination refers to the extent to which these
information sharing interactions among team members are effectively orchestrated.
More highly coordinated teams are able to Operate smoothly and efficiently without a
Substantial amount of process loss. More uncoordinated teams are characterized by
wasted motions, inefficiency, and a lack of organization in necessary operations.

In the present research, subordinates will have direct access to an equal
number of informational cues, but they will need to communicate with one another in
order to have complete information about their areas of expertise. (Note that this is a
definitional characteristic of distributed expertise.) The leader will also have direct
access to some, but not all informational cues. The leader, however, will know
which team members have direct access to which informational cues. The
communication structure will be such that each team member can communicate with
all others.

Given the experimental configuration and the requirements of distributed
expertise, the only way a team leader can obtain all necessary information is through
interaction with other members. The more efficiently the leader Obtains the

information, the greater the opportunity to use it in making a decision. That is, given

98
a fixed amount of time to make the team’s decision, the leader who obtains the

relevant information more quickly will have a greater amount of time to contemplate
and process that information. Teams that communicate with minimal inefficiency,
providing the leader with necessary information as quickly as possible can be thought
of as mm. In this context, coordination is expected to be positively related to
the team leader’s perceptions of control.
Additional Issues Relevant to the Team Context

Before concluding the discussion of risky decision making in a team context,
two additional issues merit mention. The first concerns the potential for the decision
making process to vary in the team context. The second deals with the relative
magnitude of effects in the individual and team contexts, an issue dealt with poorly in
the existing literature.
wilting

Cue weighting refers to the relative importance assigned to different cues in
making a decision. Cue weighting at the individual level is an interesting issue in
itself. In fact, a substantial literature conducted under the rubric of ”policy
eapturing" is concerned with this issue. Individual decision makers may differentially
weight various informational cues. However, one weighting in the team context is
likely to be a much more complex and significant process. Team leaders have two
potential sets of cues to consider in making a decision, task specific informational
cues (i.e., the raw data) and recommendation cues (i.e., subordinate team members’

judgments). The relative weights assigned to the two sets of cues in the leader’s

99
decision are likely to be influenced by whether the decision frame represents a loss or

a gain.

In order to more fully understand the way in which frames might affect
weighting, it is important to understand the basics of attribution theory. In general,
people prefer to take personal responsibility for success and attribute failure to
external forces (Miller & Ross, 1975; Zuckerman, 1979). Gain frames present
opportunities for success; they offer a chance to gain and no chance of loss. LOSS
frames, on the other hand, present opportunities for failure; there is no chance of gain
only the potential for loss. In the decision making context, then, it is hypothesized
that a team leader Should be more willing to take personal responsibility for decisions
under positive frames and more likely to defer responsibility for decisions made under
negative frames.

The leader may be able to defer responsibility by basing the team’s decision on
the judgments or recommendations of subordinate team members. By weighting
others’ judgments more heavily than other potential decision cues (e. g. , the raw data,
leader’s personal preference), the leader defers responsibility for the decision and
resulting outcome, diffusing it among subordinate team members. Thus, under
negative frames, leaders may be more likely to base decisions on subordinate
members’ judgments.

The Vroom-Yetton model of leader decision making also supports this notion.
When subordinate acceptance is important and unlikely the model recommends using

a completely participative process, which essentially means that decision responsibility

100
is transferred to subordinates, and the leader accepts the decision they provide. If a

team incurs a loss, as is possible under a loss frame, acceptance is both important and
unlikely. Thus, a leader may safely make decisions autocratically provided that the
decisions are correct and losses are not incurred by the team. However, given the
potential for loss coupled with the threat created by that potential (Jackson & Dutton,
1988), a leader is expected to defer responsibility and use subordinate judgments.
n' ' ff

Using the polarization argument described in the risky shift literature, some
have contended that the effects of framing are likely to be more pronounced in a team
context (e.g., Whyte, 1989). As discussed in the section on risky Shift, there are
numerous problems inherent to a within groups comparison of the riskiness of
individual and team decisions (see Figure 4). Given those problems attempting a
within groups comparison using teams seems inadvisable. The comparison could,
however, be made across the broad categories of individual and team decision
making. Since teams are expected to exhibit the same tendencies as individuals in
more extreme form, the prediction would be for greater risk aversion in gain
conditions and greater risk seeking in loss conditions. In the present Study, team
leaders are predicted to be more risk averse under gain frames and more risk seeking
under loss frames than individual decision makers. This type of comparison has not

been made in the existing literature.

101
Hypotheses for Risky Decision Making in a Team Context

The risky decision making of leaders in a team context was considered using
an application of the individual model. The nature of the linkages among variables
were largely unaltered in moving from the individual to the team context. However,
considering risky decision making in the team context required considering factors
uniquely applicable to teams. The three variables predicted to moderate the
relationship between cognitive frames and decision riskiness are shown in Figure 7.
These linkages constitute the formal team context model. Figure 8 depicts the model
with the predictors of perceived control included. It Should be understood that these
predictors are considered more speculative than the core model. Also notice the
superscript indicating the level of measurement attached to each variable in the model.
Variables were either measured at the decision level (D) or at the individual level (I)
(i.e., team leader). Below, specific hypotheses based on the preceding discussion are
derived.

The core of the model shown in Figure 7 consists of a direct linkage from
cognitive decision frame to decision riskiness. There is some suggestion that the
effects of frames predicted by prospect theory are at least as likely for teams as for
individuals (e.g., Whyte, 1989). Thus, the following hypothesis is suggested:

H4: Decision frames will have a direct effect on team leader decision riskiness.

Under conditions where alternatives are equal in subjective expected utility and

unequal in risk, team leaders presented with a gain frame will be more likely

to select a less risky alternative than those presented with a loss frame. Under

102

28:30 Eco“ c 5

9.5.2: .5333 CB: Lo .cooE 3.8235 6 959...

 

 

a xmzm ZOE—Own

 

 

:

 

 

1. wamzmzwmzoo

 

 

 

 

HzOE:.=an:_‘ mo“. ommz

 

 

 

 

 

IIIIIIIQAOEFZOO ow>.m0mwn_

 

1i“

 

 

 

 

 

m2<mm ZOE—Own
H

 

 

 

ozo:.<z_omooo
a 20.25.25sz

a >::o_ms_< 20:9sz:08.".
o >::e_ms_< use 292.252.
3.35.: moz<2m0amma
HSewage: v65.

305200 do 960..

 

 

103

the same conditions, team leaders presented with a loss frame will be more

likely to select a riskier alternative than those presented with a gain frame.

Perceptions of control were predicted to moderate the relationship between
cognitive decision frames and decision riskiness in both the individual and team
contexts. However, the moderating linkage was expected to be more difficult to
detect in the team context. The following hypothesis describes the nature of the
moderating effect.

H5: Perceptions of control will moderate the relationship between decision frame
and decision riskiness for team leaders’ decisions. Specifically, risk seeking
will be greater under loss frames than under gain frames when perceptions of
control are lower. When perceptions of control are higher, there are no
predicted differences for decision riskiness between gain and loss frames.
Two additional variables were expected to moderate the relationship between

cognitive decision frame and decision riskiness. The first variable, need for

affiliation, is a personal characteristic of the team leader. The second variable,
cohesiveness, is a team characteristic. The nature of the predicted moderator effects
are detailed in the following hypothesis:

H6: Across decisions, the leader’s need for affiliation and the team’s cohesiveness
will moderate the relationship between decision frame and decision riskiness
for a team leader’s decisions. Specifically, risk seeking will be greater under
loss frames than under gain frames when the need for affiliation and team

cohesiveness are higher than when the need for affiliation and team

104

cohesiveness are lower. When the need for affiliation and team cohesiveness
are lower, there are no predicted differences for decision riskiness between
gain and loss frames.

The same variables predicted to influence the perceived control of individual

decision makers were also expected to impact the perceived control of team leaders.

Locus of control, task knowledge, performance history, and cue ambiguity, were

proposed to affect team leaders’ perceptions of control. In addition, Figure 8 Shows a

direct linkage from centralization and coordination to perceived control that was not

relevant at the individual level.

H7:

Across decisions, team leaders with a more internal locus of control, greater
task knowledge, and a more positive performance history will have higher
perceptions of control than team leaders with a more external locus of control,
less task knowledge, and a more negative performance history. For specific
decisions, team leaders with less informational cue ambiguity, less
recommendation cue ambiguity, greater centralization, and greater coordination
will have higher perceived control than team leaders with greater informational
cue ambiguity, greater recommendation cue ambiguity, less centralization, and
less coordination.

Two additional propositions were developed for team leader risky decision

making. The first concerns decision making processes of leaders. In the team

context, leaders have the potential to use two types of decision cues: informational

and recommendation cues. Since loss frames have the potential to be perceived as

105
more threatening, negative, and undesirable (Jackson & Dutton, 1988; Miller & Ross,

1975 ; Zuckerman, 1979), the following hypothesis was derived:
H8: Under loss frames, team leaders are more likely to base decisions on
subordinate recommendations than under gain frames.

The second proposition describes the relative magnitude of cognitive framing
effects in the individual and team contexts. The present research will not make a
within groups comparison given the numerous problems with that approach. Instead,
team leaders will be compared to individual decision makers to test the following
hypothesis.

H9: The effects of decision frame on decision riskiness are likely to be stronger for
team leaders than for individuals.

The next section describes an empirical study designed and conducted to test
the hypotheses developed here. The study is meant to provide an initial test of some
of the relationships described by the individual and team context models of risky

decision making.

METHOD
E . .

Participants were 94 male and 140 female undergraduate students recruited
from two introductory management courses. Participation was voluntary, and
students received course credit for taking part in the study. In addition, all
participants had the potential to earn up to $15.00 based upon performance on the
decision making task. The average amount of money actually earned during the
experiment was approximately $10.50 per person.

Design

The research was conducted using a 2 (gain frame, loss frame) X 2
(ambiguous cues, unambiguous cues) X 20 (perceived control) repeated measures
design. This design was applied to both the individual and team contexts. In each
setting, cue ambiguity and trials were within subjects variables. Framing was not
manipulated as a within subjects variable due to evidence that the effects of an initial
frame persist, even after the frame is changed (e.g., Levin, Johnson, & Davis, 1987;
Loke, 1989). In total, the study included 41 experimental teams (20 under a loss
frame and 21 under a gain frame) and 70 individual decision makers (35 in each
framing condition).
misign 11251:

The decision making task employed in the research was an interactive,
computerized, naval command and control simulation called TIDE2 (see Hollenbeck,

Sego, Ilgen, & Major, 1991 for a detailed description). The context for the scenario

106

107

is a naval carrier team consisting of four stations. Each team member role plays the
commanding officer of a different station. The primary tasks of the team are to
monitor the airspace over their location for unidentified aircraft, evaluate the level of
threat that each aircraft represents, and respond appropriately to the aircraft based on
the threat level. An unidentified aircraft is described by nine attributes. These are:
speed, direction, range, altitude, radar, angle, size, corridor status, and identification
fried or foe (IFF). The level of threat represented by an aircraft is based upon the
configuration of these attributes. Once an unidentified aircraft appears in the
airspace, participants in the Simulation have three minutes to measure the attributes
and decide how to respond to the aircraft. The three potential responses range from
ignoring friendly aircraft that are non—threatening to defending against hostile enemy
aircraft by shooting them down. Participants are provided with rules that describe the
meanings of the attributes and their values (see Appendix A). The rules also explain
how cues are combined to determine the level of threat an aircraft represents. The
simulation can be set up for one to four players so it can be used to study both
individual and team decision making. The. team and individual versions are described
in more detail later.

A trial or game in the simulation begins when an unidentified aircraft enters
the airspace. The presence of an aircraft is indicated by a beeping noise and a red
asterisk in the middle of the computer screen. In addition, a digital clock begins to
count down the time available before a decision about how to respond to the aircraft

must be made. In the present study, each trial lasted 180 seconds, and a decision had

108
to be made before the clock reached zero. Participants (i.e., both teams and

individuals) were presented with 30 trials.
WW. As was mentioned, nine
attributes about the aircraft (labelled cues) provided the information needed to reach a
decision about the aircraft’s level of threat. The instructions to participants, Shown in
Appendix A, described the range of values for each one and how to interpret the level
of threat for each cue. The manner in which cues had to be combined to make a
correct decision was established a priori and communicated to the participants as a set
of ”rules” in the training they received on the simulation. If one considers the nine
cues independent variables .in a regression equation predicting threat, each cue could
assume a value ranging from zero (i.e., a non-threatening value) to two (i.e., a very
threatening value). In addition, the set of rules defined the main effect and
interactions among cues that determined the correct decision. In the present research,
the equation consisted of one main effect and four interactions. Table 3 lists the nine
cues and provides the equation used to determine the correct decision a priori. As
Shown in the equation in Table 3, the four interactions received unit weights and the
one main effect for IFF had a weight of two. The equation also shows which specific
cues composed each interaction (i.e. , speed and direction, altitude and corridor status,
size and radar type, angle and range). The specific cue values and cue interactions,
as provided to participants, are provided in the set of role instructions in Appendix A.
Since the cue values ranged from zero to two, the true score ranged from 0 to 20.

Ranges of true score values were associated with levels of threat. True scores

109
Table 3

I.- r- run 4..., -. ..... ..- :--r. u arr-I. - up. a...
M W
(#1) Spwd
(#2) Altitude
(#3) Size
(#4) Angle
(#5) IFF
(#6) Direction
(#7) Corridor Status
(#8) Radar Type

(#9) Range

Level of Threat = 2‘[II'5]b + l[ll'1][llf6]c + 1[#2][#7]+ l[#3][#8] + 1[#4][#9]

'Weight for cue.
l’Cue (cue #5 or IFF in example).
cInteraction between two cues (between one #1 and cue #6 or speed and direction in

example.

l 10
ranging from 0 to 10 were labelled ”Ignore," and those ranging from 11 to 20 were

labelled a ”Defend. " Participants were instructed to choose Ignore as the response to
unidentified aircraft they viewed as non-threatening and Defend as the response to
unidentified aircraft they felt were threatening. In addition, participants also had the
option of selecting a response labelled "Warn” that was a moderate response between
Ignore and Defend. More will be said about the selection of Warn relative to Ignore
or Defend later. Note for now, however, that although Warn was a W,
Warn never represented a W presented as one of the aircraft that entered
the airspace during the course of the study’s 30 trials. Participants were instructed as
to this fact in their training.

Mack. Performance feedback for each decision (i.e. , every trial) was
provided to all participants, but the feedback was delayed by one trial. For instance,
feedback for Trial 10 was provided after Trial 11. To determine the performance
outcome to be fed back, the decision registered by the individual or team was
subtracted from the correct decision. There were three possible decisions (i.e. ,
participant responses) so the performance outcome ranged from -2 to +2. An
outcome of zero indicated that a correct decision had been made. Since the Sign of
the outcome is not important in the present research, it is sufficient to note that an
absolute value of one meant that the decision was Off by one on the scale and an
absolute value of 2 meant that the decision was off by two. Figure 9 Shows the
potential performance outcomes for all combinations of correct and actual decisions.

Recall that participants had to choose among three responses: (1) Ignore, (2) Warn,

111

Actual Decision

 

 

 

 

 

 

 

Correct (1) Ignore (2) Warn (3) Defend
Decision
(1) Ignore 0 1 2
(3) Defend 2 1 O
O - HIT
1 - MISS
2 - DISASTER

Figure 9: Performance outcomes

112

and (3) Defend. The true score or correct response was always either Ignore or
Defend. From the chart in Figure 9, one can see that when the correct and actual
decisions were the same, the outcome was zero (see the upper right and lower left
boxes). Selecting a response completely Opposite of the true score (i.e. , Ignore when
Defend was correct; Defend when Ignore was correct) yielded a performance outcome
of two. An actual decision of Warn always resulted in a performance outcome of
one. In the simulation, the numerical performance outcomes were represented by the
labels Hit (0), Miss ( 1), and Disaster (2) for feedback purposes. A Hit meant that the
decision was completely appropriate (i.e. a zero). A Miss meant that the decision
was somewhat too passive or somewhat too aggressive (i.e. , a one). A Disaster
meant that the reaction was completely inappropriate and that the opposite action
should have been taken (e.g. , Ignore was selected when Defend was the correct
response; a two).
S’m l . C E . l S i

The TIDE2 Simulation is a very flexible program in that a number of structural
parameters can be altered to suit particular research needs. For instance, the program
allows you to Specify who has access to what informational cues. It also allows you
to determine who may communicate with whom. The specific configurations for the
team and individual versions of the simulation used in the present research are
described below.

Individual yersign. In the individual version, the decision making task was

performed by single subjects. The participants were stationed at a computer terminal

1 13
where they requested information about the unidentified aircraft that appeared in each

decision trial. (Recall that there were 30 trials in the present study.) Individual
participants only had to execute two functions, labelled Measure and Judgment in the
TIDE2 Simulation. Measure means directly accessing an informational cue (e. g. , size,
speed, direction, etc.) by calling it up on the computer screen. An individual
participant was able to request all nine cues. After cues were measured, it was up to
the individual to combine appropriately the cues using the provided rules. Once the
participant felt ready to make a decision and before time ran out, a Judgment had to
be made. That is, the subject had to register a decision of Ignore, Warn, or Defend.
Individual participants received performance feedback for their decisions one trial
later.

W. In the team version of the simulation, four participants
interacted in a network of four computers. Each computer was a separate work
station. The team leader was called the Carrier station, and the person assigned that
role was the team leader. The three subordinate stations were the Coastal Air
Defense (CAD), the Cruiser, and the AWACS. TIDE2 was configured so that each
team member was able to communicate with all others. However, in order to
simulate a team with distributed expertise, each member of the team was given direct
access to only five of the nine informational cues. That is, each team member could
request five cues from the computer’s file without asking other team members for the
information. Table 4 shows which team members had direct access to which

informational cues. Each subordinate team member had direct access to one unique

114

 

Table 4

by...) :M; . J. ..,....1_._ .__,. ' _ -, L“ .".,o hr. I” .11.!-
Informational Cue Station(s)

( 1) Speed: CAD(a) Cruiser Carrier(d)

(2) Altitude: CAD Cruiser(c)

(3) Size: CAD

(4) Angle: CAD AWACS(b) Carrier

(5) IFF: CAD AWACS

(6) Direction: AWACS

(7) Corridor Status: AWACS Cruiser Carrier

(8) Radar Type: . AWACS Cruiser

(9) Range: Cruiser

115

Table 4 (continued)

(a)

(b)

(c)

(d)

- [a

s .r‘. $401114”: :I'R ‘. L). LI: _ .' ._ -'n «.4101: ‘1

SPEED and DIRECTION go together, so that fast targets coming straight in are
most threatening. Speed alone and direction alone mean nothing. There is
nothing to fear if fast targets are not headed toward the group. There is nothing
to fear from Objects headed directly for the group that are moving slowly.

ANGLE and RANGE go together, so that descending targets that are close are
especially threatening. Angle alone and range alone mean nothing. Descending
targets that are far away, or close targets that are on the way up are not
threatening.

SIZE and RADAR go together, so that small objects with weapons radar are
especially threatening. There is nothing to fear from small targets with weather
radar or from large targets with weapons radar.

ALTITUDE and CORRIDOR STATUS go together, so that low flying targets
that are way outside the corridor are especially threatening. Altitude alone and
corridor status alone mean nothing. There is nothing to fear from high flying
targets well outside the corridor or low flying targets in the middle of the
corridor.

116

piece of information (i.e. , could measure a cue that no other team member could
measure). In particular, only the person at the CAD station could measure Size, only
the AWACS could measure Direction, and only the Cruiser station could measure
Range. The Carrier had access to no unique informational cues, but the Carrier was
unique in that this person knew what all the other Stations could measure. Each of
the four team members was responsible for understanding one of the four interactions
among cues. Each of the rules and the station responsible for it is listed in Table 4.
(Note that each team member received instruction regarding all four interaction rules,
but was instructed to be responsible for and become expert in only one.) Since the
expertise in the team needed to be ME, team members were not allowed to
directly access both cues relevant for their interactions. For example, the CAD
station was able to access Speed directly, but not Direction. Thus, the CAD station
had to communicate with another team member in order to find out the value for
Direction (e.g. , the CAD could have gotten Direction from the AWACS).
Essentially, team members were required to share information to perform effectively.
Information could be shared using a variety of mechanisms. In each instance,
it was necessary for one station to Measure (i.e. , access it directly from the computer)
a cue before Sharing it with another station. Once an informational cue was
Measured, three different computer communication functions had the potential to be
invoked in order to Share information with other team members. The three functions

involved mm, transmitting, and mixing information. The least expedient

method of Sharing information involved four steps once a cue had been measured.

l 17
For instance, the CAD measured Speed, the AWACS requests Speed, the CAD

receives the request and transmits Speed, and the AWACS receives Speed. Each of
the functions was required to be conducted via the computer; participants were not
allowed to speak directly to each other. A more expedient example of information
sharing occurs when team members Share information without being asked. For
instance, the CAD could send Speed to the AWACS without being asked and the
AWACS could receive it. Since the time to make a decision was limited to 180
second, expedient information sharing was desirable. The simulation included an
additional communication function that allowed team members to type in and send any
sort of brief message to another team member.

In the team version of the task, each subordinate station sent a judgment
regarding the appropriate decision to the Carrier. The Carrier then made a decision
for the entire team. Each trial ended as soon as the Carrier made a decision. The
entire team received feedback on that decision, one trial later. The feedback screen
included the team’s decision (made by the Carrier), the correct decision, and the
judgments of subordinate members.

Mute

Study participants were scheduled to come to the laboratory in groups of six,
allowing for a maximum of one team and two individual sessions at one time. Upon
arrival, subjects were randomly assigned to either a team or an individual condition
and to a loss or gain frame condition. Individuals within teams were further

randomly assigned to one of four roles within each team (i.e., the decision making

118

leader or one of the three subordinate team members). Those assigned to the
individual conditions worked independently on a computerized decision making task.
Those in the team condition worked as teams on a networked version of the same
task. All subjects Signed a consent form indicating willingness to participate in the
study.

WW. Once assigned to conditions, subjects in
the individual task had 15 minutes to read instructions (see Appendix A) orienting
them to the decision making scenario (i.e., the context, the informational cues, and
the interactions among cues), their role in the scenario, and the computerized task
itself. The role instructions also included the framing manipulation. This
manipulation described how performance was determined and the monetary outcomes
associated with each decision alternative. Outcomes were presented as either potential
gains or potential losses. Questions regarding any of the introductory material were
addressed. Next, subjects completed a manipulation check for the appropriate
framing condition to ensure that they understood the implications of the decision
frame for scoring points on the computer task and earning money. Any missed
questions on the manipulation check were covered with subjects until the issue was
understood. Subjects then had the opportunity to study their role instructions for an
additional five minutes before completing a task knowledge test.

Once task knowledge tests were completed, an experimenter conducted a two-
trial interactive training session to familiarize subjects with the operation of the

computer terminals. Subjects responded to verbal instructions during the first trial.

1 19
During the second trial, they worked independently, asking questions as necessary.

At the conclusion of training, the 30 experimental decision trials began. The first ten
were considered learning trials, during which subjects were expected to become more
familiar with the simulation and more comfortable with operating the computers.
After each decision on Trials 11 through 30, subjects were asked to indicate the
probability that they had just made a correct decision. Feedback was delayed by one
trial to allow subjects the opportunity to make their ratings before learning about their
performance.

Once all 30 decision trials were finished, subjects completed individual
difference measures. Before leaving, subjects were paid on the basis of points earned
during the decision making task. Subjects received a general debriefing and were
asked not to divulge the details of the research to others. Finally, subjects were
thanked for their participation.

W. The procedure for distributed expertise teams
was very similar to the procedure for individuals. Once assigned to a gain or loss
condition, team members had 15 minutes to read instructions appropriate for orienting
them to their unique roles, the scenario, and the decision making task. Role
instructions included information regarding the informational cues each team member
could measure and the interactions among cues for which each person was responsible
(i.e., the distributed expertise pattern). The role instructions also included the
appropriate framing manipulation which described the scoring and monetary outcomes

associated with each decision alternative. Questions regarding any of the introductory

120

material were addressed before presenting a 30 minute training video. The video
explained how to execute all computer functions. Questions were entertained at the
end of the video and subjects were given an additional five minutes to look over their
role instructions.

Each team member completed a manipulation check to ensure that the framing
condition was understood. Further instruction regarding the framing condition (i.e. ,
the scoring and monetary pay-offs associated with different decision alternatives) were
provided as necessary until each team member understood the condition. All team
members then completed a task knowledge test before beginning the interactive
training.

The two—trial interactive training session allowed the team to practice computer
functions described in the training video. Subjects responded to scripted verbal
instructions covering all functions during the first trial. They worked independently
during the second trial, asking questions as necessary. At the conclusion of training,
subjects were instructed not to talk with each other, and the 30 experimental decision
trials began. Just as with individuals, the first ten trials were considered an
opportunity for learning, during which subjects could become used to the simulation
and the computer functions. After each decision on trials 11 through 30, team leaders
were asked to indicate the probability that they had just made an accurate decision for
the team. Feedback was delayed by one session to allow leaders the opportunity to

make their ratings before learning about their performance.

121

Once all 30 decision trials were completed, team members completed a set of
individual difference measures (e. g. , need for affiliation, locus of control) and a series
of team process measures (e.g. , cohesiveness, centralization). Each team member
was then paid an amount commensurate with points earned during the decision
making simulation. Teams received a general debriefing and were asked not to
discuss the research with classmates in their management course (i.e. , other potential
subjects) until the project was concluded. Team members were thanked for their
participation before leaving.

I I . l .

Framing. The study contained two framing conditions, gain and loss. Both
conditions contained the same 30 decision trials. Participants had to decide among
the Ignore, Warn, and Defend response options for each trial. Ignore and Defend
were established as risky alternatives relative to Warn, the certain alternative, in
instructions to participants. Warn was the decision alternative for which the outcome
was certain. That is, there was no risk. The outcomes associated with selecting
Ignore and Defend were contingent on performance. In other words, if the actual
decision was Ignore, participants only received a reward if Ignore was the correct
decision. The same was true for defend. Table 5 presents the outcome contingencies
for the various combinations of actual and correct decisions.

For purposes of illustrating the relative riskiness of choosing Ignore or Defend
as opposed to Warn, it is best to compare the columns in Table 5. Notice that when

Warn was chosen, the outcome was a Miss regardless of the correct decision. Thus,

Table 5

122

 

E l E . .

lame AME-m Dem
Hit Miss Disaster

Disaster Miss Hit

123

Warn was a certain choice; it always resulted in a Miss. The outcome associated with
choosing Ignore was contingent upon whether or not Ignore was the correct decision.
If Ignore was correct, then the outcome was a Hit. If Ignore was chosen, but not
correct then the outcome was a Disaster. Similar contingencies applied when the
actual decision was Defend. The outcome was a Hit if Defend was the correct
decision and a Disaster if Defend was not correct.

Pretesting was required to determine what the outcome associated with Warn
had to be relative to the outcomes offered for the correct selection of either Ignore or
Defend. In the pretest using the same experimental targets used in the actual
research, subjects were correct on 70 percent of the decision trials when using a risk
seeking strategy (i.e. , choosing either Ignore or Defend). Thus, to make the pay-offs
associated with the risk averse strategy (i.e. , selecting Warn) equivalent to the pay-
offs for a risk seeking strategy (i.e., selecting Ignore or Defend), a 7 to 10 ratio was
established for awarding points for certain versus risky alternatives. In general terms,
selecting Warn always yielded a certain outcome that had a lower value as compared
to the correct selection of either Ignore or Defend. Likewise, the correct selection of
either Ignore or Defend was always associated with higher outcomes than choosing
Warn. However, the incorrect selection of Ignore or Defend was detrimental to
monetary rewards.

The manner in which monetary rewards were determined is the key to the
framing manipulation. Gain and loss were manipulated by the way in which pay was

provided for performance. Pay was based on points and each point was worth 5

124
cents. Recall that the pay-off for selecting Warn was equal to 70 percent of the pay-

off for correctly selecting either Ignore or Defend. In the gain condition, subjects
began with zero points and earned points for performance. Whenever Warn was
selected, subjects automatically received 7 points. Subjects received 10 points for
correctly selecting either Ignore or Defend. No points were awarded if subjects
incorrectly chose Ignore or Defend. Table 6a shows the gain frame pay-offs in terms
of the points awarded per decision or trial. Warn represented the certain decision
alternative and Ignore and Defend represented risky decision alternatives. To
determine an individual’s or a team member’s total monetary outcome, final points
were multiplied by 5 cents. Thus, the maximum that could be earned was $15.00.
The equation used to determine pay in the gain condition is also provided in Table 6a.

In the loss condition, an individual or each team member was told that they
were starting the game with $15.00. They were instructed that 5 cents would be
subtracted for each point incurred and that points would be assigned as described in
Table 6b. The table indicates that 7 points were automatically accumulated for
selecting Warn. When Ignore or Defend were chosen correctly, zero points were
added to the score. When Ignore or Defend were chosen incorrectly, 10 points were
added to the score. At the end of the game, points were multiplied by 5 cents. That
amount was then subtracted from $15.00 and the difference was the received
monetary outcome. This equation is also provided in Table 6b.

Ambiguity. Cue ambiguity was manipulated in the both individual and team

simulations. Before beginning the Simulation, subjects were provided with

125
Table 6

2.!ng. 11E

6a. Gain Frame

 

 

Galatians
5 l E . .
Ignore Warn—Defend
Ignere 10 7 0
mm
Defend 0 7 10
Pay = (# points)(.05) 10 points = 50 cents 7 points = 35 cents
6b. Loss Frame
Lesst‘Iarne
‘ i i
Ignore Warn Defedd
km 0 7 10
Semi Deejsigd
Diced 10 7 0

Pay = 15.00 - (# points)(.05) 10 points = 50 cent loss 7 points = 35 cent loss

126

information about the meanings associated with ranges of cue values. For example,
they were told that speed ranged from 100 to 800 miles per hour. Cue values were
categorized as either non-threatening or very threatening. These ranges, however,
were not continuous. That is, between the two categories there were certain values
that did not fall into either range. These values were considered ambiguous; subjects
could not be certain exactly how they should be categorized. Using speed as an
example, the non-threatening range was 100 to 275 miles per hour and the threatening
range was 525 to 800 miles per hour. Thus, a speed from 276 to 524 would be
ambiguous. Across the 30 experimental trials, 15 contained ambiguous cues and 15
contained unambiguous cues. Targets with ambiguous and unambiguous cues were
randomly presented in three blocks consisting of ten trials. Each block of ten
contained five ambiguous and five unambiguous targets presented randomly within the
block.

In the team context, a second type of ambiguity was also relevant. Each
subordinate member provided a decision recommendation to the team leader. These
judgments, then, could be combined and used by the leader as an additional
informational cue (i.e. , a recommendation cue) in making the team’s final decision.
To the extent that subordinates’ judgments were consistent with each other, the
recommendation cue was unambiguous. However, the cue became more ambiguous
with greater disagreement. For each experimental decision trial, the three
recommendation cues were coded and combined post has to form a scale indicating

the level of ambiguity in the set of recommendation cues. For instance, when all

127

three subordinates suggested the same decision, the recommendation one was coded

" l ," representing the lowest possible ambiguity. When only two subordinates’
judgments were the same, the recommendation cue was coded "2,” representing
moderate ambiguity. Finally, when all three subordinates offered different judgments,
the recommendation cue was coded "3," representing high ambiguity.

Measures

Misign risldness. The level of risk represented by a decision was assessed
by examining the decision alternative selected on experimental decision Trials 11
through 30. (Recall that Trials 1 through 10 were excluded to allow participants with
a learning period.) For both individuals and team leaders, the extreme choices (i.e.,
Ignore and Defend) were coded as risk-seeking and the middle alternative or the
certain option (i.e. , Warn) was considered risk averse.

Pergivgl genteel. Perceived control was also measured on decision Trials 11
through 30. As soon as the decision for a trial had been made, individuals and team
leaders were instructed to record their perceived probability that the decision they just
made was a correct decision. Instructions for making the probability ratings had to be
carefully articulated because if the decision just made was "Warn,” it was obviously
not correct in an objective sense because the targets were constructed so that Warn
was never the correct response. Thus, subjects were asked to rate the probability that
they would have been correct had they attempted to make a correct decision (i.e. , if
they had responded with Ignore or Defend). Subjects were told to use this strategy

even if they had actually selected the Warn option. (See Appendix B for the

128
perceived probability measure and instructions.) Thus, after making a decision on

Trials 11 through 30, individual decision makers and leaders used a rating seale
ranging from .50 to 1.0 to indicate the probability that a correct decision could have
been made on that trial. They were allowed to write in any value for perceived
probability that fell within the range from .50 to 1.0 (see Appendix B).

W. Task knowledge tests relevant to the individual simulation
and the team leader position in the team simulation were developed and used in
previous research using the TIDE2 paradigm. Those tests were used in the present
research (see Appendix C). Both tests included items related to the specific cues the
particular station could measure, the meanings of the cue values, and the interactions
among cues. The team leader’s test also included items regarding which subordinate
stations could measure which cues. Each test consisted of 13 multiple choice items.
The internal consistency reliabilities for the individual and team leader tests in this
study were .66 and .69, respectively.

W. Locus of control was measured using a 29-item, forced-
choice scale developed by Rotter ( 1966). For each of the 29 items, subjects were
asked to chose which of two statements they agreed with more. Items were coded so
that higher scale scores indicated a more internal locus of control (i.e., 1=extemal
and 2 =intemal). In previous research, split-half and test-retest reliabilities associated
with the scale ranged from .65 to .70. The alpha computed for the present research

was .77.

129
W- Performance history was measured by averaging the

decision outcomes (i.e. , performance) on the second block of experimental trials (i.e. ,
11 through 20). The first set of 10 trials was excluded because learning was expected
to occur within that block and was considered likely to make performance unstable.
Each trial was scored either zero if an incorrect response was made or 1 if a correct
decision was made. The 10 scores were then averaged to create an overall measure
of performance history.

mm. The need for affiliation of team leaders was assessed
using a scale derived from Jackson’s (1965) Personality Research Form. Jackson’s
original scale consisted of 20 true/false items. In the present research, 10 of the
seemingly more face valid items were used and rated on a four-point rating scale
ranging from 1 = "very false” to 4 = ”very true." The internal consistency
reliability for this adapted scale was .71.

QgheshLedess. Cohesiveness was measured using a three item scale previously
employed in team decision making research using the TIDE2 program. The five-point
response format for each item ranged from ”strongly agree" to "strongly disagree."
Coefficient alpha for that scale was .81.

mm. Centralization was assessed by examining the patterns of
communication during each decision trial (i.e., experimental trials 11 through 30) in
the Simulation. Specifically, the number of communications directed at the Carrier
(i.e. , the leader and decision maker) were compared to the average number of

communications directed at subordinate stations. This ratio (i.e. , Carrier

130

communication over other) provided a non-perceptual index of centralization for each
of the experimental trials.

erdinadgd. Recall that the objectively correct decision for a given trial was
determined by a regression equation involving four interactions and one main effect.
To be completely knowledgeable, a team leader Should know all five components
before making a decision. Within in the TIDE2 simulation the leader could obtain
each of these five components with varying efficiency. Thus, coordination was
operationalized as the number of components possessed by the leader (i.e. , zero to

five) divided by the number of inefficiencies involved in obtaining each component.

RESULTS

The majority of hypotheses were tested using repeated measures regression for
within and between subjects variables, as described by Cohen and Cohen (1983).
Before discussing the tests of specific hypotheses, the general data analytic strategy is
described.

The first step in a repeated measures regression analysis with between and
within subjects effects is partitioning the variance in the criterion. A heuristic
illustrating the partitioning of between and within subjects variance is presented in
Figure 10. The diagram is an adaptation from Cohen and Cohen (1983). The large
circle (Y) represents the total variance in some dependent variable, the criterion. The
circle (Y) is first bisected into the portions representing the total between and within
subjects variance. In Figure 10, the left-hand portion of the circle represents the total
between subjects variance and the right-hand portion represents the total within
subjects variance. The smaller circles each represent independent variables used to
predict some portion of (Y). Predictor (H) is a between subjects factor. Predictor (I)
is a within subjects factor. The predictor labelled (H x I) is the interaction of the
between and the within subjects predictors (H) and (I).

In partitioning the variance in (Y) into the between and within subjects
components, we can fully define variance associated with areas (m + c) and (n + d
+ e) respectively. Once the between and within subjects variance has been
determined, the R square for each predictor (i.e., H, I, and H x I) can be calculated.

The regression analysis is conducted hierarchically such that all the within subjects

131

132

Between

 

 

Y-Total variance in dependent variable
B-Between subjects variance in Y
w-Within subjects variance in Y

H-A between subjects predictor

l-A within subjects predictor

H x I-A within/between interaction

Adapted from Cohen and Cohen (1983), p. 440.

Figure 10: Partitioning the between and within
subjects variance

133

effects (i.e. , within subjects main effects and interactions involving only within
subjects variables) are entered first. The between subjects effects are entered next,
followed by interactions involving between and within subjects variables. A unique R
square can be calculated for each within subjects effect, each between subjects effect,
and each within/between interaction.

Using the area labels in Figure 10, the R square for the between subjects
predictor (H) in the diagram is defined as the ratio: c/(c + m). Notice that at this
point we are only interested in the proportion of (B) variance accounted for by (II).
This is the unique R square for the between subject effect of (H). The areas
represented by (n), (d), and (e) are not relevant for the between subjects analyses.
Later we will calculate the percentage of total (Y) variance accounted for by (H).
Focusing on the unique within subjects effects, we disregard the areas (c) and (m) and
focus instead on (d), (e) and (n). For instance, the R square for (I) is calculated
using the proportion: d/(d + e + n). The R square for the (H x I) interaction is
computed by: e/(d + e + n). The unique between and within subjects R squares can
then be tested for significance with separate F-tests. One can also assess the
proportion of total variance accounted for by a given predictor (e. g., H, I, or H x I).
In the ratio representing this type of R square, the denominator includes the total
variance instead of the unique variance associated with the between or within subjects
effects. For instance, the ratio for the total proportion of variance accounted for by
(H) is: c/(c + m + n + d + e). Similarly, the ratio for the total proportion of

variance accounted for by (I) is: d/(c + m + n + d + e). These effects can be

134
tested for significance using corresponding F-tests. See Cohen and Cohen (1983) for

a more detailed description of the data analytic strategy and the specific equations
used in computing R squares and corresponding F-tests (pages 437-448).
IesLoflndixidnalflxpntheses

Having described the general analysis strategy, it can now be applied in testing
the hypotheses developed in the present research. The individual level hypotheses are
addressed first. The means, standard deviations, and meaningful correlations among
individual level variables are presented in Table 7. As noted in the table, the n for
correlations between variables that were measured between subjects was 70 (i.e. , the
number of subjects) and the n for correlations between the within subjects variables
was 1400 (i.e., 70 multiplied by the 20 decision trials).

Hypotheses 1 and 2 address the relationship between cognitive frame,
perceived control, and risk taking. These hypotheses are restated below:

H1: Decision frames have a direct effect on decision riskiness. Under conditions
where alternatives are equal in subjective expected utility and unequal in risk,
individuals presented with a gain frame will be more likely to select a less
risky alternative than those presented with a loss frame. Under the same
conditions, individuals presented with a loss frame will be more likely to select
a riskier alternative than those presented with a gain frame.

H2: Perceived control will moderate the relationship between decision frame and
decision riskiness. Specifically, risk seeking will be greater under loss frames

than under gain frames when perceptions of control are lower. When

135

 

Table 7
11.1””: l'Ill
Variable Mean sd 1 2 3 5 7
(1) Frame
(2) Risk .89 .31 -.226*
(3) Perceived
Control 79.31 17.54 -.087* .359“
(4) Cue Ambiguity -.303 -.652*
(5) Task
Knowledge 9.34 1.98 -.159 .301* .305*
(6) Locus of
Control 1.44 .17 -.041 -.093 .022 -.131
(7) Performance .73 .14
History
(8) late Perceived
Control 80.30 7.02 .513*

Ngte. Correlations involving variables 5, 6, 7, and 8 are based on [I = 70. Correlations
involving any other variables are based on n = 1400.

136

perceptions of control are higher, there are no predicted differences for

decision riskiness between gain and loss frames.

Since H2 introduces perceived control as a moderator of the relationship
between framing and risk taking described in H1 (i.e., H2 is an elaboration of H1),
these hypotheses were tested together using repeated measures regression. In this
analysis, decision frame was a between subjects variable and perceived control was a
within subjects variable. Partitioning the variance in decision riskiness showed that
14 percent of the variance was due to between subjects factors and 86 percent was
due to within subjects factors. The results of the regression analysis are presented in
Table 8. Decision frame, perceived control, and their interaction were all significant
predictors of decision riskiness. Decision frame accounted for 26 percent of the
WW. Perceived control and the interaction between decision
frame and perceived control accounted for 15 and about 4 percent of the mm
mm, respectively. The nature of the interaction between decision frame
and decision riskiness, plotted in Figure 11, was as predicted in H2. Individuals were
generally taking risks under loss frames across levels of perceived control.
Individuals under gain frames tended to be risk averse when perceived control was
low, but risk seeking when perceived control was high. Given the significant main
effect for decision frame and the significant frame by perceived control interaction,

which was in the predicted direction, both H1 and H2 were supported. Finally, the

137

 

Dependent Variable = Decision Riskiness

 

 

 

 

 

 

 

 

 

 

 

Independent
Variables Unique R2 F—Value df Total R2
(l) Perceived

Control .15" 15.10* (19,1292) .129*
(2) Decision

Frame .26" 23.89“ (1,68) .165"
(3) (1) * (2) .035" 3.49* (19,1292) .195*

*p. < .05

wWithin subjects effect

”Between subjects effect

0.8

0.6

0.4

0.2

138

Risk

 

 

 

 

 

Gain Loss

Decision Frame

—+‘ Low Control + High Control

Figure 11: Moderating effect of perceived control

139

W in decision riskiness explained by the three variables was 20

percent.‘

Hypothesis 3 addressed potential predictors of perceived control. The
hypothesis is restated below:

H3: Across decisions, individuals with a more internal locus of control, greater
task knowledge, and a more positive performance history will have higher
perceptions of control than individuals with a more external locus of control,
less task knowledge, and a more negative performance history. For specific
decisions, individuals with less informational cue ambiguity will have higher
perceived control than individuals with greater informational cue ambiguity.
Two separate analyses were conducted for this hypothesis. The first tested the

relationship between performance history and perceived control. The second tested

the relationships between all the other predictors and perceived control. Two analyses
were required because the criterion for the analysis involving performance history was
only based on the last 10 games. The last 10 games were used, as opposed to the

entire 20, in order to examine the effects of previous performance (i.e. , in games 11—

20) on subsequent perceived control (i.e. , in games 21-30). Since only two variables

were involved, a Simple correlation between performance history and later perceived

 

‘ At first glance, the fact that 20 percent of the total variance was predicted
when it was reported that the decision frame accounted for 29 percent of the unique
variance may seem confusing. The fact that both results are possible is apparent form
Figure 10. The denominator for the total variance is the area of the entire circle (Y),
whereas the denominator for the decision frame effect is only the portion dealing with
between factors and thus, always less than (Y).

140

control was computed. As Shown in Table 7, the correlation, r=.513, was Significant
at p < .05 , supporting the predicted relationship.

The remaining components of Hypothesis 3 were tested using the repeated
measures regression for between and within subjects effects. Partitioning the total
variance in perceived control showed that 15 percent was due to between subjects
factors and 85 percent was due to within subjects factors. Cue ambiguity was the
within subjects variable in this analysis. There were two between subjects variables,
task knowledge and locus of control. Since Hypothesis 3 was designed to be
exploratory, all the two-way interactions among the potential predictors of perceived
control were also tested. The results of the repeated measures regression for
Hypothesis 3 are shown in Table 9. AS predicted, task knowledge and locus of
control were both positively related to perceptions of control. That is, individuals
with higher task knowledge and individuals with a more internal locus of control
tended to perceive greater control. Cue ambiguity was not a significant predictor of
perceived control. None of the interactions involving cue ambiguity were
significantly related to perceived control either. However, the interaction between
task knowledge and locus of control was Significantly related to perceived control
[F=7.34, df=(1,68)]. The interaction between locus of control and task knowledge
in predicting perceived control is plotted in Figure 12. The graph Shows that for
individuals with an internal locus of control, perceived control increased as task
knowledge increased. On the other hand, for individuals with an external locus of

control, perceived control tended to remain relatively constant across various levels of

141

 

Dependent Variable = Perceived Control

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Independent
Variables Unique R2 F-Value df Total R2
(1) Cue

Ambiguity .026" 2.38 (1,68) .022
(2) Task

Knowledge .297" 28.70" (1,68) .067*
(3) Locus of

Control .076" 559* (1,68) .079*
(4) (1) * (2) .034” 3.16 (1,68) .108*
(5) (1) * (3) .007w .69 (1,68) .114*
(6) (2) * (3) .097“ 7.34* (1,68) .129*

*p. < .05

"’Within subjects effect

l’Between subjects effect

142

Perceived Control

 

100

80*

 

 

 

 

so ~-

40L"

20—:

 

 

 

Task Knowledge

+External +’ Internal

Figure 12: Interaction between task knowledge and
locus of control

143

task knowledge. The total amount of variance in perceived control accounted for by

all main effects and all interactions was 13 percent. Hypothesis 3 was partially

supported.

WW
Hypotheses 4, 5, 6, and 7 all dealt with the risky decision making of leaders in

a team context. Results related to those four hypotheses are presented in this section.

Table 10 provides the means, standard deviations, and meaningful correlations among

the variables relevant to the team context. As noted in the table, the n for

correlations between variables measured between subjects was 41 (i.e. , the number of

subjects) and the n for correlations between within subjects variables was 840 (i.e., 41

multiplied by the 20 decision trials).

Hypotheses 4, 5, and 6 were all concerned with predicting the decision
riskiness of team leaders. Each hypothesis is restated below:

H4: Decision fiames will have a direct effect on team leader decision riskiness.
Under conditions where alternatives are equal in subjective expected utility and
unequal in risk, team leaders presented with a gain frame will be more likely
to select a less risky alternative than those presented with a loss frame. Under
the same conditions, team leaders presented with a loss frame will be more
likely to select a riskier alternative than those presented with a gain frame.

H5: Perceptions of control will moderate the relationship between decision frame
and decision riskiness for team leaders’ decisions. Specifically, risk seeking

will be greater under loss frames than under gain frames when perceptions of

144

 

 

Table 10
film Qentext Sergeladgg Table
E 5!. C l 'n
Variable Mean sd 1 2 3 4 5 6 7
(1) Frame
(2) Risk .86 .14 -.465*
(3) Perceived
Control 79.72 7.17 .102 .070
(4) Need for
Affiliation 3.22 .42 .061 -.013 .221
(5) Cohesiveness 3.59 .63 -.244 .096 -.009 .077
(6) Task
Knowledge 6.93 2.94 -.109 .052 .319* -.060 .346*
(7) Locus of
Control 1.43 .19 -.056 .240 -.179 .050 .053 -.062
(7) Performance
History .715 .115
(8) late Perceived
Control 81.52 8.49 -.201
i ' ' l ' n
Variable Mean sd 1 2 3 4 5 6
(1) Frame
(2) Risk .86 .34 -.196*
(3) Perceived
Control 79.6918.89 .038 .323“
(4) Cue Ambiguity -.376* -.537*
(5) Recommend.
Ambiguity 2.02 .83 .071*-.134* -.327* .402*
(6) Centralization .31 .15 -.149* .027 .026 -.084* .003
(7) Coordination .97 .98 .011 -.006 .014 -.019 -.036 .421*

Note. Between subjects correlations are based on n = 41. Within subjects correlations

arebased onn = 820. "p. < .05

H6:

145

control are lower. When perceptions of control are higher, there are no
predicted differences for decision riskiness between gain and loss frames.
Across decisions, the leader’s need for affiliation and the team’s cohesiveness
will moderate the relationship between decision frame and decision riskiness
for a team leader’s decisions. Specifically, risk seeking will be greater under
loss frames than under gain frames when the need for affiliation and team
cohesiveness are higher than when the nwd for affiliation and team
cohesiveness are lower. When the need for affiliation and team cohesiveness
are lower, there are no predicted differences for decision riskiness between
gain and loss frames.

These three hypotheses were tested together using the repeated measures

regression procedure. The results of that analysis are Shown in Table 11. Overall,

the main effects and interactions involved in Hypotheses 4, 5, and 6 accounted for

almost 17 percent of the variance in the decision riskiness of team leaders.

Partitioning the variance in decision riskiness showed that 7 percent of the variance

was due to between subjects factors and 93 percent was due to within subjects factors.

Hypothesis 4 predicted a main effect for decision frame on decision riskiness.

Table 11 Shows that decision frame accounted for 63 percent of the unique between

subjects variance in decision riskiness. Examining the beta weight associated with

decision frame also shows that the relationship was in the predicted direction (i.e. ,

leaders were more risk taking under loss frames). Thus, Hypothesis 4 was supported.

Table 11

 

146

Dependent Variable = Decision Riskiness

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Independent
Variables Unique R2 F-Value df Total R2
(1) Perceived
Control .110" 5.30"' (19,741) .103*
(2) Decision
Frame .630b 6634* (1,39) .146“
(3) (1) e (2) .009" .45 (19,741) .154*
(4) Need for
Affiliation .021" .82 (1,39) .155*
(5) Team
Cohesiveness .007" .29 (1,39) . 156*
(6) (2) * (4) .000” .00 (1,39) .156*
I (7) (2) e (5) .142” , 6.43* (1,39) .166*
*p. < .05

wWithin subjects effect

l’Between subjects effect

 

147

Hypothesis 5 predicted an interaction between decision frame and perceived
control in predicting decision riskiness. AS Shown in Table 11, perceived control did
have a main effect on decision riskiness, accounting for 11 percent of the unique
within subjects variance, but the hypothesized interaction with decision frame was not
Significant. Thus, Hypothesis 5 was not supported.

Finally, Hypothesis 6 predicted that the team leader’s nwd for affiliation and
the team’s cohesiveness would each interact with decision frame in predicting decision
riskiness. Only the interaction between decision frame and team cohesiveness was
significant, accounting for 14 percent of the unique between subjects variance. To
determine if the nature of the interaction was as predicted, the interaction between
decision frame and team cohesiveness was plotted in Figure 13. It was expected that
the greatest differences between risk taking under gain and loss frames would be
exhibited when team cohesiveness was high. AS shown in Figure 13, the opposite
result was obtained. The greatest differences in risk taking under loss and gain
frames occurred under conditions of low team cohesiveness. More specifically, risk
seeking was greater under loss frames than under gain frames when team
cohesiveness was low. When team cohesiveness was high, risk taking was more
comparable under gain and loss frames. It is interesting to note that under loss
frames, decision riskiness decreased as team cohesiveness increased, while under gain
frames riskiness increased as team cohesiveness increases. Hypothesis 6 was not

supported.

0.8

0.6

0.4

0.2

148

Risk

 

 

 

 

 

Gain Loss
Decision Frame

+ Low Cohesiveness + High Cohesiveness

Figure 13: Moderating effect of team cohesiveness

149

Predictors of team leaders’ perceived control were considered in Hypothesis 7.
That Hypothesis is presented below:

H7: Across decisions, team leaders with a more internal locus of control, greater
task knowledge, and a more positive performance history will have higher
perceptions of control than team leaders with a more external locus of control,
less task knowledge, and a more negative performance history. For specific
decisions, team leaders with less informational cue ambiguity, less
recommendation cue ambiguity, greater centralization, and greater coordination
will have higher perceived control than team leaders with greater informational
cue ambiguity, greater recommendation cue ambiguity, less centralization, and
less coordination.

Two separate analyses were conducted for this hypothesis. The first tested the
relationship between the team’s performance history and leader’s perceived control.
The second tested the relationships between all the other predictors and the leader’s
perceived control. Two analyses were necessary because the criterion for the analysis
involving the team’s performance history was only based on the last 10 games. The
last 10 games were used, as opposed to the entire 20, in order to examine the effects
of previous performance (i.e. , in games 11-20) on subsequent perceived control (i.e. ,
in games 21-30). (Recall that the same procedure was used at the individual level.)
Since only two variables were involved, a simple correlation between performance

history and later perceived control was computed. AS shown in Table 10, the

150

correlation, r=—.201 , was not significant. Thus, the predicted relationship was not
supported.

The remaining components of Hypothesis 7 were tested using the repeated
measures regression for between and within subjects effects. Partitioning the total
variance in team leaders’ perceived control showed that 14 percent was due to
between subjects factors and 86 percent was due to within subjects factors. In this
analysis, there were four within subjects variables: cue ambiguity, recommendation
ambiguity, coordination, and centralization. Task knowledge and locus of control
were the between subjects variables. The more exploratory two-way interactions
among the potential predictors of perceived control were excluded because there was
not sufficient power to test them. The results of the repeated measures regression for
Hypothesis 7 are shown in Table 12. Overall, the six predictors accounted for 37
percent of the total variance in perceived control. However, cue ambiguity was by
far the largest contributor, accounting for almost 40 percent of the within subjects
variance. In fact, one ambiguity was the only significant predictor of perceived
control in the analysis. Thus, only one of the hypothesized relationships in H7 was
supported.

Pr ' D i i n 'n

Hypothesis 8 was concerned with the risky decision making processes of team

leaders. In particular, H8 dealt with use of informational cues and recommendation

cues under loss and gain frames. The Specific hypothesis is restated below:

151

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 12
r i n R l f r i 7
Dependent Variable = Perceived Control
Independent
Variables Unique R2 F-Value df Total R2
II ( 1) Cue
Ambiguity .395" 34.46* (15,585) .340*
(2) Recom’d.
Ambiguity .012“’ 1.05 (15,585) .350’“
(3) Coord. .000”’ .00 (1,39) .351*
(4) Centrality .006" .51 (1,39) .356*
(5) Task
Knowledge .011” .44 (1,39) .365“
(6) Locus of
Control .006” .22 (1,39) .370*
"p. < .05

"Within subjects effect

bBetween subjects effect

 

152

H8: Under loss frames, team leaders are more likely to base decisions on
subordinate recommendations than under gain frames.

This hypothesis called for a comparison of R squares obtained under gain and
loss frames. Since decision frame was a between subjects variable, this means that R
squares obtained using two unique samples needed to be compared. Since there is
currently no direct test for comparing R squares obtained from two different samples
(Cohen & Cohen, 1983), an alternative approach was employed.

As a first step toward testing Hypothesis 8, two separate regression analyses
were conducted, one for leaders under gain frames and one for leaders under loss
frames. In each analysis, the three subordinate recommendations were entered as
predictors of the team leader’s decision. In order for H8 to be potentially supported,
the R square obtained for loss frame leaders had to be larger than the R square
obtained for gain frame leaders. Regression results Showed that the R square for
leaders under loss frames was .536 and the R square for leaders under gain frames
was .431. Both R squares were significant.

Simply comparing the R squares and concluding that the one obtained under
loss frame conditions was larger offered only weak support for Hypothesis 8,
especially Since both R squares were significant. Even though there is no direct test
for determining whether or not two R squares obtained from unique samples are
Significantly different (Cohen and Cohen, 1983), additional steps were taken to
determine if the two R squares were substantially different. The first step involved

obtaining the confidence interval for the gain frame R square. The second step was

153

to ascertain whether or not the R square for the loss frame R square fell within that
interval.

The confidence interval for the gain frame R square (.431) ranged from .337
to .525. Thus, the loss frame R square of .536 fell outside this interval. This finding
further supported Hypothesis 8, showing that team leaders under loss frames are

considerably more likely to rely on subordinate recommendations in making team

decisions than team leaders under gain frames.

 

Hypothesis 9 dealt with a comparison of the effects of decision frames on
decision riskiness in the individual and team contexts. The specific prediction is
restated below:

H9: The effects of decision frame on decision riskiness are likely to be stronger for
team leaders than for individuals.

A hierarchical regression analysis with dummy coding was used to test this
hypothesis. All of the variables involved were between subjects factors. In the
analysis individuals were dummy coded zero and team leaders were dummy coded
one. The results of the hierarchical regression are presented in Table 13. Only the
first step in the regression analysis was significant, essentially demonstrating a main
effect for decision frame that was already established in tests of Hypotheses 1 and 4.
Since framing effects obtained for individual decision makers did not differ
significantly from framing effects evidenced for team leaders, Hypothesis 9 was not

supported.

154

Table 13

W
Dependent Variable = Decision Riskiness

 

 

 

 

 

Independent
Variables Unique R2 F-Value df Total R2
(1) Decision

Frame .276 41.62“ (1,109) .276*
(2) Individual/

Leader .009 1.42 (1,109) .286*
(3) (1) * (2) .000 .04 (1,109) .286*

 

 

 

 

 

 

*p. < .05

DISCUSSION

The primary goal of this research was to explore the effects of cognitive
frames on risky decision making in the individual and team contexts. More
specifically, the research was designed to study potential moderators of the
relationship between cognitive decision frames (i.e. , loss and gain frames) and
decision risk purported by prospect theory. Nine formal hypotheses were developed,
three that dealt with individual level issues, five concerned with a leader in a team
context, and one involving a comparison of the individual and team contexts. The
empirical tests of these hypotheses were presented in the previous section. To
understand more fully the findings obtained in this study, this section begins with a
summary and interpretation of the results. The chapter concludes with a discussion of
the primary contributions and limitations of the research.
Wits

W. At the individual level, results showed that
cognitive decision frames had the expected effects on decision risk. Findings
supported Hypothesis 1 which predicted greater risk seeking under loss frames and
greater risk aversion under gain frames. These results are consistent with prospect
theory propositions and the empirical results of the original empirical research
conducted in developing prospect theory (Kahneman & Tversky, 1979, 1984; Tversky
& Kahneman, 1981). The framing effects detected in this study can be thought of as
an extension of previous research given the substantial differences between the

decision task employed in the present research and the types of tasks typically used in

155

156
prospect theory research. Paper and pencil hypothetical choice problems have been

the most usual type of decision task employed in cognitive framing research prompted
by prospect theory. By comparison, it seems likely that the information processing
task employed in the present research (i.e. , the TIDE2 simulation) represented a more
complex and engaging decision scenario. Previous research by Schurr (1987)
obtained results consistent with prospect theory for individuals working on a complex
bargaining task. However, that work was attempting to study team level phenomena.

A complex decision task was employed in the present research in an attempt to
create some variability in perceived control. A key theme in the present research was
that cognitive framing effects (i.e. , risk seeking under loss frames and risk aversion
under gain frames) were likely to be a function of the low perceived control generated
by the lottery-type tasks typical of prospect theory research. It was argued here that
perceived control is a critical boundary condition of the effects of cognitive decision
frames on decision riskiness. In particular, it was argued in Hypothesis 2 that under
gain frames, individuals with high perceived control would be more risk seeking than
individuals with low perceived control. Individuals under loss frames were expected
to be more consistently risk seeking, regardless of the level of perceived control.
Results showed the expected effect for perceived control. Thus, it was concluded that
framing effects are more pronounced under conditions in which individuals perceive
low control and are less likely to be evidenced under conditions of high perceived

control.

157

In Hypothesis 3, an attempt was made to identify some of the predictors of
individual decision makers’ perceived control. Across decisions, cue ambiguity, task
knowledge, and locus of control were considered. Since the hypothesis was
considered more speculative, the two-way interactions among the predictors were also
considered. Results showed that task knowledge and locus of control were both
significant predictors of perceived control in decision making settings. Individuals
with greater task knowledge and individuals with a more internal locus of control
perceived greater control than individuals with less task knowledge and individuals
with a more external locus of control.

In addition, the interaction between task knowledge and locus of control was
significant. Results showed that the effects of task knowledge on perceived control
were greatest for individuals with an internal locus of control. In other words,
internals with low task knowledge perceived less control than internals with high task
knowledge, and slightly less than extemals with low task knowledge. Internals with
high task knowledge perceived the greatest control. Individuals with an external locus
of control had a similar level of perceived control across levels of task knowledge.
This finding is not surprising given the definition of the locus of control construct.
By definition, those with a more internal locus of control are more likely to view
outcomes and events as contingent on the self, while those with a more external locus
of control attribute causality to external forces (Rotter, 1966). Since one’s level of
task knowledge is a personal characteristic, it seems reasonable that task knowledge

would have a greater impact on perceived control for internals. According to the

158
theory, extemals do not believe they have much control regardless of personal

abilities like task knowledge.

An individual’s performance history on the decision task was expected to be
positively related to perceived control. The results supported this hypothesis with a
significant correlation between early task performance (i.e. , trials 11-20) and later
perceived control (i.e. , trials 21-30). Individuals who made more accurate decisions
early in the simulation had greater perceived control toward the end of the session. A
positive performance history was expected to be interpreted as reinforcing feedback
thereby increasing perceptions of control.

In general, the individual level hypotheses were supported by the empirical
results. The effects of cognitive frames on decision riskiness were demonstrated
using a complex decision task. In addition, results showed that perceived control
moderated the relationship between framing and decision risk in the expected fashion.
Three of the four hypothesized predictors of perceived control, performance history,
task knowledge, and locus of control, were substantiated. An interpretable interaction
between task knowledge and locus of control in the prediction of perceived control
was also detected.

W. Prospect theory framing effects were predicted for leaders of
decision making teams in Hypothesis 4. As expected, results showed that leaders
under gain frames tended to make risk averse decisions while leaders under loss
frames made more risk seeking decisions. As at the individual level, perceived

control was expected to moderate the relationship between decision frame and

159

decision riskiness for team leaders. However, for team leaders, the interaction was
not significant. Hypothesis 5 was not supported. Nonetheless, perceived control did
have a significant main effect on decision riskiness. Leaders with greater perceived
control tended to make riskier decisions than leaders with lower perceived control.

In Hypothesis 6, two additional moderators of the relationship between
decision frame and decision riskiness were proposed in the team context: the team
leader’s need for affiliation and the team’s cohesiveness. In this sample, there was
very little variance in the need for affiliation of team leaders. Thus, the results did
not support the need for affiliation moderator prediction. Furthermore, the interaction
between decision frame and need for affiliation was not significant. The interaction
between decision frame and team cohesiveness was significant. However, the nature
of the interaction was not as predicted. Leaders of highly cohesive teams were
expected to make more risk seeking decisions under loss frames and more risk averse
decisions under gain frames. Leaders of less cohesive teams, on the other hand, were
predicted to make more consistently risk seeking decisions. The obtained interaction
is depicted in Figure 13. (The predicted interaction is shown in Figure 6.)

leaders of highly cohesive teams were more risk seeking under loss frames
than under gain frames. However, compared to leaders of less cohesive teams,
leaders of highly cohesive teams were relatively more; risk seeking under gain frames.
Examining the interaction, it seems that the cognitive framing effect was simply more
pronounced for leaders of teams low in cohesiveness. These leaders tended to be

more risk seeking under loss frames and more risk averse under gain frames. The

160

leaders of more highly cohesive teams tended to be more consistent in the riskiness of
their decisions across loss and gain frames. Rather that exacerbating the effects of
framing on decision riskiness, cohesiveness appeared to temper the effects of
cognitive decision frames. If the typical effects of cognitive frames on decision risk
represent nonrational decision making, then cohesiveness would appear to urge team
leaders toward greater rationality in risky decision making. It was assumed that team
leaders in highly cohesive teams would be more influenced by decision frames
because they would perceive greater social costs for failing to gain (i.e. , under gain
frames) and losing (i.e. , under loss frames). Previous research has shown that
cohesiveness fosters an atmosphere of greater social support (e.g. , Griffith, 1989).
The present results seem to indicate that instead of focusing on the potential social
costs, leaders may have perceived greater " social license" to disregard the decision
frames as the result of team cohesiveness.

Hypothesis 7 was considered more exploratory, dealing with predictors of team
leaders’ perceived control. As at the individual level, performance history, cue
ambiguity, task knowledge, and locus of control were considered. In addition, the
relationships between perceived control and recommendation ambiguity, coordination,
and centrality were also studied. Only the relationship between cue ambiguity and
perceived control was significant. The more ambiguous the informational cues, the
less control leaders perceived. The effect was quite large, accounting for nearly 40
percent of the within subjects variance and approximately 34 percent of the total

variance. Since cue ambiguity and recommendation ambiguity were highly correlated

161
(r=.40, p < .01), it seems reasonable that only the first predictor entered into the

repeated measures regression for perceived control (i.e. , cue ambiguity) was
signifieant.

Task knowledge and locus of control, both significant predictors of perceived
control for individual decision makers, were not predictive of team leaders’ perceived
control. It may be that in a team context, individual difference variables related to
the team leader or any one individual have less of an impact than at the individual
level. It was argued that risky decision making in the team context would be
qualitatively different from individual risky decision making, in that perceived control
would be a more complex issue in the team context. It appears that, in this case, the
individual model for predicting perceived control was not very useful when applied to
the team context.

However, even the more team-specific predictors of perceived control were not
significant in the repeated measures regression analysis. Neither coordination or
centrality were predictive of team leaders’ perceived control. Examining the zero-
order correlations, we find that coordination and centralization were significantly
related to each other (r=.42, p. < .01), but not to perceived control. It also seems
that the stability of the coordination measure was questionable in that the standard
deviation (.98) was comparable to the mean (.97).

Measurement problems aside, one potential reason that perceived control was
not predicted by coordination and centralization is that team leaders did not have a

sense of the extent to which their teams were coordinated and centralized. Both

162

centralization and coordination were objective measures derived from the simulation
itself by assessing relevant communication patterns. It may have been difficult for
team leader’s to obtain an accurate picture of the communication patterns in their own
teams and interpret them in a meaningful way. The TIDE2 simulation was configured
such that team leaders could only track communication involving themselves. Thus,
while they may have been able to track communication patterns in which they were
included, leaders could not readily evaluate the extent of communication among
subordinates. So, for instance, a team leader may not have had a clear idea of the
amount of communication directed toward him or her, relative to the amount of
communication among subordinate stations (i.e. , centralization). Likewise, leaders in
the simulation could not directly ascertain the efficiency of communication patterns
among subordinates, one aspect of coordination. These difficulties in assessing
communication patterns may have been exacerbated by the fact that each decision trial
only lasted 3 minutes. In addition, even if leaders were aware of the levels of
coordination and centralization in their own teams, those levels may not have been
meaningful in the absence of some comparative standard. That is, without knowing
the levels of coordination and centralization in other teams, it seems that the levels in
a leader’s own team may have been more difficult to interpret.

Although there was a significant and positive relationship between performance
history and perceived control at the individual level, the same relationship predicted
for team leaders was not substantiated. The correlation between performance on trials

11—20 and perceived control on trials 21-30 was not significant and in the wrong

163

direction (i.e. , team leaders with a more positive performance history perceived less
control). It may be that, compared to individual decision makers acting alone, team
leaders perceived less of a linkage between themselves personally and decision
performance. Thus, while individual decision makers who performed well reported
greater perceived control over subsequent performance, the lack of correlation in the
team context suggests that leaders did not to make as strong a connection between
previous performance and subsequent control.

The final team leader hypothesis addressed the manner in which leaders make
risky decisions under gain and loss frames. It was expected that leaders would rely
more on recommendation cues under loss frames than under gain frames. The
reasoning being that leaders would be more likely to attempt to diffuse responsibility
for potential losses than for potential gains. The results supported this prediction.
The R square for recommendation cues in predicting risky decisions under the loss
frame was outside the confidence interval for the R square obtained under the gain
frame. This test was somewhat indirect, but offered some initial support for
Hypothesis 8, nonetheless.

WWW Hypothesis 9 Predicted that
the effects of cognitive decision frames on decision riskiness would be stronger for
team leaders than for individual decision makers. This hypothesis was not supported.
The hierarchical regression results showed only a main effect for framing, the effect
that was obtained in both the individual and team contexts in tests of Hypotheses 2

and 5, respectively. The results obtained here indicate that framing effects were not

164

substantially stronger for team leaders in a team context. The contention most
commonly expressed in the literature is that the effect of cognitive frames on risky
decision making are likely to be exacerbated for teams relative to individuals (e.g.
Whyte, 1989). Although the literature is not clear on this point, the prediction may
be more relevant for different types of decision making teams, such as teams that
reach decisions by consensus. In the teams used for the present research, individual
team leaders were ultimately responsible for making the team decision. Thus, it
appears that the team context does not dramatically alter the effects of cognitive
frames on decision riskiness for individual decision makers.

Summary. Results were generally supportive of the study’s central
hypotheses. Cognitive decision fiames had the predicted effects on decision riskiness
for individual decision makers and leaders of decision making teams. In addition,
certain factors were found to moderate the relationship between decision frames and
decision riskiness. For individual decision makers that moderator was perceived
control. Individuals under gain frames tended to make more risk averse decisions
when perceived control was low and more risk seeking decisions when perceived
control was high. Individuals under loss frames made more consistently risk seeking
decisions regardless of perceived control. For team leaders, team cohesiveness
moderated the relationship between decision frame and decision riskiness. However,
the nature of that relationship was not as expected. Instead of exacerbating the effects
of cognitive frames, high cohesiveness tended to diminish the effects of cognitive

decision frames. Overall, then, the study was successful in demonstrating cognitive

165
framing effects for individuals and team leaders and in defining boundary conditions

for those effects.

In this section, the major contributions and limitations of the present research
are considered. The discussion centers primarily around three themes: control, risky
decision making in the team context, and decision accuracy.

W. Considerable attention has been devoted to understanding
risky decision making processes. In an attempt to explain the departures from
rationality often evidenced in decision making under conditions of risk, Kahneman
and Tversky (1979) developed prospect theory. Empirical research to date has largely
supported the frame dependent shifts in risk preference predicted by prospect theory
(Kahneman & Tversky, 1979; 1984; Tversky & Kahneman, 1981). Essentially,
previous work has shown that individuals tend to be risk seeking when a risky
decision is fiamed as a loss and risk averse when the decision is framed as a gain.

The present research sought to explore the boundary conditions of prospect
theory’s framing effects. Although Kahneman and Tversky (1984) have contended
that cognitive framing effects are both "robust” and ”pervasive, " previous research
has demonstrated predicted effects in only limited contexts, for certain types of
decision problems. Framing effects have typically been examined using hypothetical
choice problems (e.g., Fagley & Miller, 1987; 1990; Kahneman & Tversky, 1979;
1984) or gambling tasks (e.g., Tversky & Kahneman, 1981). Both types of decision

making scenarios involve lottery-type tasks characterized by low control.

166

The issue of context is critical because it may have a direct bearing on
perceived control. Clearly, not all risky decision making occurs in a gambling or
lottery-type context. In particular, risky decisions within the organizational
environment are likely to differ substantially from the gambling scenario. For
instance, organizational decision makers have expertise, knowledge, and a variety of
other criteria on which to base their risky decisions. Presumably in this type of
context, decision makers can assess their own probabilities of making a correct
decision. In other words, the decision maker can figure his or her own ”odds” and
respond accordingly. In a lottery context, the probabilities of making a correct
decision are given. Since a lottery is a purely chance event, there is little reason for
a decision maker to believe that the applicable probabilities of making a correct
decision are different from those that are given. This type of scenario can be
contrasted with a situation in which the outcome for an accurate risky decision is
based on something other than chance. When the probability of making a correct
decision is dependent on some personal or situational factors, individuals may assess
their own personal probabilities.

The present study demonstrated the predicted moderating effect for perceived
control. Individuals and team leaders who reported lower perceived control made
decisions in accordance with prospect theory predictions. In contrast, those reporting
higher perceived control made more risk seeking decisions regardless of the cognitive
decision frame. The complex decision making simulation allowed for the

manipulation of certainty through varying the ambiguity of informational decision

167

cues. The use of the TIDE2 simulation represents an extension of previous research
which primarily used paper and pencil hypothetical choice problems. Even though
the simulation was an artificial task in which study’s participants would not normally
be engaged, participants’ outcomes were real, not hypothetical. Rewards were
actually contingent upon the decisions made by the individuals and team leaders.

This study’s findings may be interpreted in terms of their implieations for the
role of subjective expected utility (SEU) in risky decision making. One might argue
that under conditions of low control, SEU does not affect a decision maker’s choice.
However, under conditions of high control SEU may be the determinant of risk
taking. In other words, a decision maker is simply more likely to take a risk when
SEU is greater. This interpretation would be consistent with the study’s results since
perceived control was operationalized as the probability of making a correct decision.

In any case, the present findings suggest that "nonrationality" in risky decision
making may not be as pervasive or as problematic as suggested in the literature (e.g. ,
Kahneman & Tversky, 1984). Although the present research is preliminary, it does
seem to suggest that outside a lottery context, decision makers base risky decisions on
personal perceptions of control instead of relying exclusively on frame in which the
decision problem is presented.

Previous research has demonstrated that perceived control is positively related
to a number of desirable organizational outcomes. The results of Spector’s (1986)
meta-analysis indicated that perceived control is positively related to job satisfaction,

organizational commitment, job involvement, performance, and motivation. The

168

results also demonstrated that perceived control is negatively related to stress,
absenteeism, intentions to turnover, and turnover. The decision making literature has
theorized that perceived control would be positively related to risk taking (Baird &
Thomas, 1985; Vlek & Stallen, 1980). It has also been argued that risk taking would
be greater for skill-dependent decision tasks than for chance—dependent tasks (Vlek &
Stallen, 1980). The results of the present research seem to support these theoretical
contentions.

Future research should focus on the further development of the control
construct in the risky decision making context. If perceived control is to have solid
implications beyond the SEU interpretation, then it is important to demonstrate that
”probability of success" is only one aspect of the perceived control construct. For
example, one could apply Averill’s (1973) typology of control to the decision making
context, defining and exploring the effects of behavioral, cognitive, and
informational control on risk taking.

Future research should also devote further attention to the predictors of
perceptions of control. In the present research, task knowledge and locus of control
were predictive of perceptions of control. Decision task familiarity or experience
may also be predictive of perceptions of control. This potential predictor of control
seems particularly relevant in an organizational context. Organizational decision
makers will vary in terms of their experience making risky decisions. More

experienced decision makers should have enhanced perceived control.

169
W. Although little scholarly research has

addressed risky decision making in the context of teams, the topic is a significant one.
In organizations, a great deal of risky decision making occurs in a team context.
Hackman (1990) provided several examples of decision making teams in
organizational settings, noting both their criticality and their shortcomings. Ilgen et
al. (in press) have commented on the importance of decision making teams in
organizations, noting the paucity of relevant empirical research currently in existence.
Since organizations are likely to continue to rely on teams to make important
decisions, the topic is one that needs additional theoretical and empirical attention.

In terms of the team context, research concerning the effects of cognitive
frames on decision risk is at the embryonic stage. This study attempted to extend and
apply what is known at the individual level to the team context. Simply
demonstrating the effects of cognitive frames on decision risk in a team context
represents a reasonable contribution to the literature at this point. The present study
showed that cognitive frames affect the decision riskiness of leaders operating in the
context of a hierarchical team with distributed expertise. A reasonable next step in
the development of team research might be an attempt to replicate these findings using
different types of teams (e.g., democratic and consensus seeking). A successful
replication would extend our knowledge of the team phenomenon, while failure to
replicate would help further define the boundaries of the team context effects.
Regardless of the results, a meaningful contribution to the literature could be gleaned

from research involving teams of varying structures.

170

Additional team research may also continue to explore the boundary conditions
of framing effects by considering potential moderators of the relationship between
cognitive frames and decision riskiness. The present study found that team
cohesiveness vitiated the effects of cognitive frames. Cohesiveness is an indicator of
the quality of the relationships within the team in terms of how attracted team
members are to the team. Other indicators of team quality may similarly moderate
framing effects. For instance, teams that rate high on collectivism (i.e. , the tendency
to be group oriented) and/or team member exchange (i.e. , effective relationships
among team members) may be less subject to the effects of cognitive decision frames.

W. Studying the effects of cognitive decision frames using
information processing tasks provides a unique opportunity to consider the issue of
decision accuracy. Part of the difficulty with lottery tasks and gambling scenarios is
that there are no objectively "right" or "wrong” decisions. An information processing
task has an objectively correct decision that is established a priori. Thus, future
research may examine not only the effects of decision fiames on risk taking, but also
their effects on decision accuracy.

This study’s findings provide some basis for speculating that perceived control
may have an impact on the relationship between decision frame and decision
accuracy, similar to its effect on the relationship between frame and risk. This study
demonstrated that risk is consistently high under loss frames, regardless of the level
of perceived control. This implies that not much "differential” decision making

occurs under loss frames; risky options are consistently preferred. In contrast, under

I
L
" F
33 ,
t
l.

~'l
a

171

gain frames individuals make differential decisions on the basis of perceived control.
They discriminate by taking less risk under conditions of low control and more risk
under conditions of high control. One could hypothesize that decision accuracy is
consistent across levels of perceived control under loss fiames, but that decision
accuracy is greater when perceived control is greater under gain frames. Given the
significant theoretical and practical implications of decision accuracy, this hypothesis,
or at least this issue, warrants attention in future research.
Cnnfludinafiemarks

Decision making under conditions of risk is an important theoretical and
applied topic. The study of the effects of cognitive decision frames on risk appears to
be a particularly interesting area of research in which the potential for progress
related to both individuals and teams is great. The present research demonstrated
framing effects for individuals working on a complex decision task. The study also
made some important preliminary steps toward defining the boundaries of framing
effects. Perceived control seems to be a critical determinant of risky decision making
at the individual level, particularly as a moderator of the cognitive framing effects
predicted by prospect theory. This research has also made some initial steps toward
identifying predictors of perceived control in the risky decision making context.

The present study has extended the individual literature by demonstrating
cognitive framing effects for leaders in a team context. Results showed that
individual difference variables did not have significant effects in the team context,

neither as moderators of the relationship between cognitive frames and decision risk,

172

nor as predictors of perceived control. However, team characteristics may be
important in defining the boundaries of framing effects in the team context. In the
present research, team cohesiveness moderated prospect theory framing effects.
Although the effects of cognitive frames may not be as pervasive and all
encompassing as previously indicated in the literature (e. g. , Kahneman & Tversky,
1984), that) does not mean the topic is any less important as an area of research. On
the contrary, defining the boundaries of prospect theory’s fiaming effects and
considering the implications for decision accuracy, especially within an organizational
context, may be a particularly interesting and important area of research. Given the
prevalence of teams in organizational settings (e.g., Hackman, 1990; Ilgen et al., in
press), extending individual level findings and determining their applicability to the

teams may be an equally worthwhile endeavor.

APPENDICES

APPENDIX A

APPENDIX A

GENERAL INSTRUCTIONS

V TEA

W

The year is 1994 and you are a part of a US naval carrier group’s command and
control team, stationed in the Middle East. A regional conflict between two nations
in this area has recently broken out. Your mission is to protect sea-going commercial
traffic in the area from accidental or intentional attacks. As history indicates, this is a
highly sensitive task. For example, in 1987, an Iraqi jet accidentally fired two Exocet
missiles into the Frigate U.S.S. Stark, killing 37 American servicemen and crippling
the vessel. One year later, the U.S.S. Cruiser Vincennes accidentally shot down an
Iranian passenger plane killing 290 innocent civilians. Any repeat of mistakes of this
kind will probably lead to a withdrawal of American forces from the area. Such a
withdrawal would have disastrous economic and political ramifications that would
spread well beyond this region.

W

A naval carrier battle team is an awesome array of ships and support units. It
consists of a concentric ring of missile firing warships which protect the aircraft
carrier at its center. The aircraft carrier, in return provides an overall umbrella of air
protection for the entire task force. The carrier’s 90 planes can unleash air strikes
against targets at land, sea and even under water. A carrier group can dominate up to
196,000 square miles of Ocean. Your carrier group consists of the Carrier itself, a
Ticonderoga class Aegis Cruiser, AWACs reconnaissance planes and a land based
Coastal Air Defense (CAD) unit. Although the Carrier itself is equipped with some
air patrol capacities, the Cruisers, AWACS and CAD units provide the bulk of air
traffic patrol. Taken together, the air patrol groups on the Carrier, the Cruiser, the
AWACs and the CAD unit make up the command and control team.

W

The team of which you are a part, will role play the Commanding Officers of various
units in the carrier group. Your mission is to monitor the air space surrounding the
carrier group, making sure that neutral ships are not attacked. In performing this role,
you must make certain that you do not allow loss of life resulting from accidental or
intentional attacks on ships in the task force. At the same time, it is also of
paramount importance that you do not inadvertently shoot down friendly military
aircraft or any civilian aircraft. Many passenger aircraft fly through the region, and
friendly military aircraft from nations not involved in the conflict also patrol the area.
The navy can ill-afford any mistakes of either the Stark or Vincennes variety.

173

174

W

There are four roles in this simulation. The leader is the Commanding Officer (C0)
of the Aircraft Carrier. The other team members include the C0 of the AWACS air
reconnaissance planes; the C0 of the Aegis Cruisers, and the C0 of the Coastal Air
Defense (CAD) unit. The team’s task is to decide what response the carrier group
should make toward incoming aircraft. Aircraft that are being tracked on radar are
called targets. Teams base their decisions on data they collect by measuring
characteristics of the air targets. These measures are obtained from sophisticated
radar equipment. The team must make a critical choice regarding each target. There
are three potential responses, IGNORE, WARN, and DEFEND. These are described

below:

IGNORE:

DEFEND:

This means that the carrier group should devote no further attention to
the target and instead focus on other possible targets in the area. The
group should never ignore a target that might possibly attack. This
would most assuredly lead to loss of lives on the ship attacked.

This means that the carrier group sends a message to the target
identifying the group and alerting the target. Warn will NEVER BE
THE ABSOLUTELY CORRECT RESPONSE. Since in reality all
targets are either threatening or non-threatening, WARN will never be
the most appropriate response. The decision to WARN, however, is
not as bad as selecting DEFEND when the correct response is IGNORE
or as bad as selecting IGNORE when the correct response is DEFEND.

This is ”weapons away” and means to attack the target with Tomahawk
cruise missiles. A defend decision cannot be aborted once initiated.
Defend is an appropriate response when the team feels attack is
imminent.

175

W

The incoming air targets can be measured on nine attributes. These are listed below
along with the ranges of possible values on the attributes:

(1) Speed:

(2) Altitude:

(3) Size:

(4) Angle:

(5) IFF:

(6) Direction:

(7) Corridor
Status:

(8) Radar Type:

(9) Range:

100 to 800 miles per hour (mph)

5,000 to 35,000 feet

size of the target ranging from 15 to 50 meters

-15 (rapid descent) to +15 degrees (rapid ascent)

”Identification Friend or Foe. " This is a radio signal that
identifies whether an aircraft is civilian, para-military, or
military, ranging from .2 Mhz (an airliner) to 1.6 Mhz (a
fighter).

from +40 degrees (passing far to the east or west of the
Carrier) to 00 degrees (coming straight in to the Carrier)

a corridor is a 20 mile lane open to commercial air
traffic. Status is expressed in terms of miles from the
center of the corridor, ranging from 1 mile (in the middle
of it) to 50 miles (way out of it)

the kind of radar possessed by the aircraft ranging from
Class 1 (weather radar only) to Class 9 (weapons radar)

distance of the aircraft from the Carrier ranging
anywhere from 20 to 200 miles

176

 

D 1.1 ullullk Ill, _, __ _ O .LLL- '

In general, the degree to which an incoming target is threatening depends on its
standing on these nine attributes. There are five simple rules to remember in
determining the danger associated with any target:

 

(a) All else equal, in terms of IFF, W are more threatening than
civilian targets (see attribute #5)

(b) SPEED and DIRECTION go together, so that W
are most threatening (see #l—#6 above). Spwd alone and direction alone
mean nothing. There is nothing to fear if fast targets are not headed toward
the group. There is nothing to fear from objects headed directly for the group
that are moving slowly.

(e) ANGLE and RANGE go together. so that W
are especially threatening (see #4-#9) above. Angle alone and range alone
mean nothing. Descending targets that are far away, or close targets that are
on the way up are not threatening.

(d) ALTITUDE and CORRIDOR STATUS go together, so that W
W are eSPeeially threatening (see #247
above). Altitude alone and corridor status alone mean nothing. There is
nothing to fear from high flying targets well outside the corridor or low flying
targets in the middle of the corridor.

(e) SIZE and RADAR go together, so that W are
especially threatening (see #3—#8 above). There is nothing to fear from small
targets with weather radar or from large targets with weapons radar.

 

for example, if the team detected an (a) military aircraft that was (b) flying in straight
and fast, (c) was close and descending, (d) was flying low and way outside the
corridor, and (e) was small and had weapons radar; a hostile attack would be
indicated and the team should DEFEND.

If the team detected (a) a civilian aircraft, that is (b) passing slow at an angle, (c) was
far away and ascending, (d) was flying high and in the middle of the corridor and (e)
was large and had weather radar; this would indicate a passenger plane that should be
IGNORED.

Whether the appropriate response is DEFEND or IGNORE will not always be
obvious. Sometimes, a target will appear threatening according to some of the rules
but not all. The team will have to combine the rules in order to make a correct
decision. Sometimes the team may just decide to WARN a target, rather than risk
incorrectly choosing IGNORE or DEFEND. Remember that all the rules count
equally in determining what judgment should be made. The team’s expertise and
experience will be important for handling the non-obvious targets.

177

W115]:
The Commanding Officer of the Carrier is the team leader and the person who

ultimately decides what to do for each target. The other team members, however,
also make decisions about how to handle the targets. These decisions are sent to the
Carrier as recommendations. Each team member has expertise that is unique to his or
her role. That expertise comes in the form of the person’s (a) ability to measure
attributes and translate raw data into judgments regarding threat, and (b) the
person’s knowledge of the rules.

For example, although all team members know that military aircraft are more
threatening than non-military aircraft, only two people in the team can actually
measure this characteristic of the target. In addition, only these two players will be
trained to know exactly how raw data on IFF (i.e., radio signal Mhz.) can be
translated into terms of "non-threatening” or "very threatening."

Similarly, each member of the team will have to W of the four
combination rules (e. g. , memorize how speed and direction go together). Thus, at
least one member of each team will be an expert on each of the four combination
rules.

The C0 of the Carrier also memorizes one rule, but he or she can only measure three
attributes. The distinctive competency of the Carrier, however, is that its CO knows
what all the other team members are experts in.

RAW

Patrol sessions refer to the time that your team is responsible for monitoring air
traffic in your designated area. While you are monitoring traffic, you will be
stationed at a computer monitor. This monitor will have four icons on it. These are
there to remind you that there are four members on the team: the Carrier (i.e. , the
team leader), the AWACS (i.e., plane), the Coastal Air Defense unit (i.e., the land
mass) and the Aegis Cruiser (i.e., the ship). A red dot in the middle of the screen
indicates the presence of a target in your airspace. A clock on the screen tells how
long before the C0 of the Carrier must render a decision. The target will begin to
blink and beep at an increasing rate when there is less than 30 seconds left to
respond. If the Leader fails to make any decision with respect to the target, this will
be recorded as a NO CALL and treated as if the team decided to IGNORE it.

The bar at the top of the screen indicates what you can do while on patrol. There are
basically five things that you can do when trying to make a decision about a target:
Measure, Query, Receive, Transmit, and make a Judgment. Each of these functions
will either be described in the videotape or in your training sessions.

178
W

After every trial except the first, team members will receive a report telling them how
well the team is performing. W. This means that the
results of Trial 1 will be provided after Trial 2 and the results of Trial 2 will be
shown after Trial 3, etc. Each feedback screen informs the team of their decision, as
well as the "correct decision. " There are three possible outcomes from an encounter.
The team’s total effectiveness and monetary rewards will be expressed in terms of
points associated with each outcome.

HIT:
A hit means that the team’s decision was exactly correct. This means that a
friendly or civilian aircraft was appropriately IGNORED or that your team
correctly DEFENDED against a hostile military aircraft. mg
in ’ v 1 .

MISS:
Your team will incur a ""miss whenever WARN is chosen as the team’s
response. Remember that WARN rs never the best response, but it is better

than selecting IGNORE or DEFEND incorrectly. W
DISASTER:

A disaster means that the team’s decision was incorrect. That is, the team
decision was IGNORE when it should have been DEFEND or; the team’s
decision was DEFEND when the target should have been IGNORED. A
disaster means that one of the ships in the group was struck by a missile (if
overly passive), or that one of the ships shot down a friendly target (if overly

aggressive) WW

179
FRAMING MANIPULATION S

MW
At the end of the simulation the team’ s POINTS GAINED will be totaled. mm

W15. So, a hit 1s worth 50 cents, a miss is worth 35 cents, and a disaster
is worth nothing. Hits are worth 30 percent more money than misses because they
are 30 percent more difficult to obtain. (The 30 percent estimate is based on the
AVERAGE performance of previous teams across trials.)

The feedback screen provides information on the judgments rendered by each station
and the team. (Remember that feedback is delayed by one trial.) The team’s
performance history is recorded in terms of the total number of Hits, Misses, and
Disasters acquired. The team’s total points are also recorded on the feedback screen.
There is also a space for a goal, but no specific goal has been assigned to your team.
The fwdback screen also shows a projection of the team’s total score based on what
the final total will be at the end of the entire session if performance continues at the
current level of proficiency. To figure out how much money each member of your
team will receive, multiply the total points by .05. For example, if the number of
total points at the end of the session is 200, each team member would receive $10.00.

You will be given more information about earning money before you begin the
simulation.

W

At the end of the simulation the team’s POINTS LOST will be totaled. mm
W. So, for a hit the team loses nothing, a miss
means a 35 cent loss, and a disaster means a 50 cent loss. The deduction for misses
is 30 percent less than the deduction for disasters because disasters are 30 percent
more difficult to avoid. (The 30 percent estimate is based on the AVERAGE
performance of previous teams across trials.)

The fwdback screen provides information on the judgments rendered by each station
and the team. (Remember that feedback is delayed by one trial.) The team’s
performance history is recorded in terms of the total number of Hits, Misses, and
Disasters acquired. The team’s total points lost are also recorded on the feedback
screen. There is also a space for a goal, but no specific goal has been assigned to
your team. The feedback screen also shows a projection of the team’s total score
based on what the final total will be at the end of the entire session if performance
continues at the current level of proficiency. To figure out how much money each
member of your team will receive, multiply the total points by .05 and subtract that
amount from $15.00. For example, assume that the number of total points at the end
of the session is 100. You would multiply 100 by .05 for a product of $5.00.
Subtract $5.00 from $15.00. Each team member would receive $10.00.

You will be given more information about earning money before you begin the
simulation.

180
W

821W. The Carrier is the leader who makes the team’s final
decision. The carrier can only measure and interpret three things (1) Speed, (2)
Angle, and (3) Corridor Status. The range of values and degree of threat associated
with each are shown below.

 

Degree of Threat
Non-Threatening Very Threatening
Speed 100 to 275 mph 525 to 800 mph
Angle +15 to +8 degrees -8 to -15 degrees
Corridor Status 0 to 8 miles out 22 to 50 miles out

 

The leader is unique in knowing the areas of expertise for all other areas. The table
below shows which attributes each station can measure.

CAD X X X X X
AWAC X X X X X
Cruiser X X X X X

 

Summary of How to Determine Threat Levels

(a) All else equal, in terms of IFF, W are more threatening than
civilian targets.

(1)) SPEED and DIRECTION go together, so that W
are most threatening. Speed alone and direction alone mean nothing. There is

nothing to fear if fast targets are not headed toward the group. There is
nothing to fear from objects headed directly for the group that are moving
slowly.

(e) ANGLE and RANGE go together. so that W
are especially threatening. Angle alone and range alone mean nothing.

Descending targets that are far away, or close targets that are on the way up
are not threatening.

181

(d) ALTITUDE and CORRIDOR STATUS go together, so that W

that am way gutside the corridor are especially threatening. Altitude alone
and corridor status alone mean nothing. There is nothing to fear from high

flying targets well outside the corridor or low flying targets in the middle of
the corridor. *****

(6) SIZE and RADAR go together. so that mm are
especially threatening. There is nothing to fear from small targets with

weather radar or from large targets with weapons radar. *****

***** Carrier must memorize this rule!!

182
Wale
W. The Coastal Air Defense (CAD) unit is a specialist in
the measurement and interpretation of five target attributes, (1) speed, (2) altitude, (3)
size, (4) angle, and (5) IFF.

The range of values and degree of threat associated with each are shown below.

 

Degree of Threat
N on-Threatening Very Threatening
Speed 100 to 275 mph 525 to 800 mph
Altitude 35,000 to 27,000 ft 13,000 to 5,000 ft
Size 50 to 40 meters 20 to 10 meters
Angle + 15 to +8 degrees -8 to -15 degrees
IFF .2 to .6Mhz 1.4 to 1.8Mhz

 

Summary of How to Determine Threat Levels

(a) All else equal, in terms of IFF, W are more threatening than
civilian targets.

0)) SPEED and DIRECTION go together, so that W

are most threatening. Speed alone and direction alone mean nothing. There is
nothing to fear if fast targets are not headed toward the group. There is
nothing to fear from objects headed directly for the group that are moving
slowly.*****

(e) ANGLE and RANGE go together. so that W
are especially threatening. Angle alone and range alone mean nothing.

Descending targets that are far away, or close targets that are on the way up
are not threatening.

(d) ALTITUDE and CORRIDOR STATUS go together, so that W
thatmmmtsidetmm are esrreeially threatening Altitude alone
and corridor status alone mean nothing. There is nothing to fear from high
flying targets well outside the corridor or low flying targets in the middle of
the corridor.

183
(e) SIZE and RADAR go together. so that MW: are

especially threatening. There is nothing to fear from small targets with
weather radar or from large targets with weapons radar.

***** CAD must memorize this rulell

184
W301:
BMW. The AWACs unit is a specialist in the measurement
and interpretation of five target attributes, (1) angle, (2) IFF, (3) direction, (4)
Corridor status, and (5) Radar Type.

The range of values and degree of threat associated with each are shown below.

 

Degree of Threat
N on-Threatening Very Threatening
Angle +15 to +8 degrees -8 to -15 degrees
IFF .2 to .6Mhz 1.4 to 1.8Mhz
Direction 40 to 22 degrees 08 to 00 degrees
Corridor Status 0 to 8 miles out 22 to 50 miles out
Radar Type Classes 1&2 Classes 8&9

 

Summary of How to Determine Threat Levels

(a) All else equal, in terms of IFF, W are more threatening than
civilian targets.
0’) SPEED and DIRECTION go together, so that fast 181.2915 gaming :1 . I I .

are most threatening. Speed alone and direction alone mean nothing. There is
nothing to fear if fast targets are not headed toward the group. There is
nothing to fear from objects headed directly for the group that are moving
slowly.

(e) ANGLE and RANGE go together. so that 11W
are especially threatening. Angle alone and range alone mean nothing.

Descending targets that are far away, or close targets that are on the way up
are not threatening.*****

(d) ALTITUDE and CORRIDOR STATUS go together, so that W
that age way eateige the eerridor are especially threatening. Altitude alone
and corridor status alone mean nothing. There is nothing to fear from high
flying targets well outside the corridor or low flying targets in the middle of
the corridor.

185

(e) SIZE and RADAR go together. so that mm are
especially threatening. There is nothing to fear from small targets with

weather radar or from large targets with weapons radar.

***** AWACs must memorize this rule!!

186

W

W. The Aegis Cruiser is a specialist in the measurement
and interpretation of five target attributes, (l) Corridor status, (2) Radar Type, (3)

Range, (4) Speed, and (5) Altitude.

 

 

Degree of Threat

N on-Threatening Very Threatening
Corridor Status 0 to 8 miles out 22 to 50 miles out
RadarType Classesl&2 Classes8&9
Range 200 to 110 miles 60 to 20 miles
Speed 100 to 275 mph 525 to 800 mph
Altitude 35,000 to 27,000 ft 13,000 to 5,000 ft

Summary of How to Determine Threat Levels
(a) All else equal, in terms of IFF, W are more threatening than
civilian targets.

(b) SPEED and DIRECTION go together, so that WW

(0)

(d)

are most threatening. Speed alone and direction alone mean nothing. There is
nothing to fear if fast targets are not headed toward the group. There is
nothing to fear from objects headed directly for the group that are moving
slowly.

ANGLE and RANGE go together. so that W
are especially threatening. Angle alone and range alone mean nothing.

Descending targets that are far away, or close targets that are on the way up
are not threatening.

ALTITUDE and CORRIDOR STATUS go together, so that Wm
W are especially threatening. Altitude alone
and corridor status alone mean nothing. There is nothing to fear from high
flying targets well outside the corridor or low flying targets in the middle of
the corridor.

187
(e) SIZE and RADAR go together. so that MW are

especially threatening. There is nothing to fear from small targets with
weather radar or from large targets with weapons radar. *****

***** Cruiser must memorize this rule!!

APPENDIX B

APPENDIX B

PERCEIVED CONTROL
Probability of the Correct Decision

At chance levels, the probability of being correct if you select Ignore or Defend is .50
(i.e., a 50 percent chance of getting a HIT). Based on the performance of previous
teams, the average team working on average targets has a .70 probability of being
correct if Ignore or Defend is selected (i.e., a 70 percent chance of getting a HIT).

If you were 100 percent certain of being correct, the probability of getting a HIT
would be 1.0.

Beginning with trial 11, please indicate the probability that you would have received a
HIT for choosing Ignore or Defend. Even if you selected WARN, please make the
rating based on what you believe your probability of being correct would have been
had you selected Ignore or Defend.

Your rating can range from .50 to 1.0.

 

 

.50 .70 1.0
PROBABILITY PROBABILITY FOR PROBABILITY
AT CHANCE LEVELS AVERAGE TEAM AT CERTAINT Y
WORKING ON AN
AVERAGE TARGET

During the feedback screen following each of the listed trials, record the
probability that you would have been correct had either Ignore or Defend been
your response.

 

 

 

 

 

 

 

 

 

 

Trial 11: probability Trial 21: probability
Trial 12: probability Trial 22: probability
Trial 13: probability Trial 23: probability
Trial 14: probability Trial 24: probability
Trial 15: probability Trial 25: probability
Trial 16: probability Trial 26: probability

 

188

 

Trial 17:
Trial 18:
Trial 19:

Trial 20:

 

 

 

 

probability
probability
probability

probability

189

Trial 27:
Trial 28:
Trial 29:

Trial 30:

 

 

 

 

probability
probability
probability

probability

 

APPENDIX C

APPENDIX C

TASK KNOWLEDGE TESTS

Indium

1

1.5 Mhz represents which of the following?

A non-threatening target.

A very threatening target.

A target on the border between two threat levels.
A target that is out of the possible range.

O-OC'N

+20 degrees of angle represents which of the following?
a A non-threatening target.
b A very threatening target.
c A target on the border between two threat levels.
d A target that is out of the possible range.

5 miles out side the corridor represents which of the following?
A non-threatening target.

A very threatening target.

A target on the border between two threat levels.
A target that is out of the possible range.

Q0 0"”

35 degrees of direction represents which of the following?
a A non-threatening target.
b A very threatening target.
c A target on the border between two threat levels.
(1 A target that is out of the possible range.

30,000 foot altitude represents which of the following?

A non-threatening target.

A very threatening target.

A target on the border between two threat levels.
A target that is out of the possible range.

(DO-5"”

20 meters represents which of the following?
a A non-threatening target.
b A very threatening target.
c A target on the border between two threat levels.
d A target that is out of the possible range.

190

10

11

12

13

191

850 miles per hour represents which of the following?

A non-threatening target.

A very threatening target.

A target on the border between two threat levels.
A target that is out of the possible range.

O-OO‘N

Class 1 radar represents which of the following?
a A non-threatening target.
b A very threatening target.
c A target on the border between two threat levels.
d A target that is out of the possible range.

80 miles for range represents which of the following?
a A non-threatening target.
b A very threatening target.
c A target on the border between two threat levels.
d A target that is out of the possible range.

Which of the following combinations represents a more threatening target?
Slow and land radar targets.

b I-Iigh flying and fast targets.

c Descending and inside the traffic corridor targets.

d Outside the traffic corridor and high flying targets.

e Close and descending targets.

3,

Which of the following combinations represents a more threatening target?
Slow targets with weather radar.

b High flying and fast targets.

c Descending and inside the traffic corridor targets.

d Small targets with weapons radar.

e Close and ascending targets.

3)

Which of the following combinations represents a more threatening target?
Slow and land radar targets.

b High flying and fast targets.

c Descending and inside the traffic corridor targets.

d

e

9)

Fast targets headed toward the group.
Close and ascending targets.

Which of the following combinations represent a more threatening target?
a Slow and land radar targets.
b High flying and fast targets.
c Descending and inside the traffic corridor targets.
d Outside the traffic corridor and low flying targets.
e Close and ascending targets.

192
Carrier
1 200 miles per hour represents which of the following?

a A non-threatening target.
b A very threatening target.
c A target on the border between two threat levels.
d A target that is out of the possible range.
2 25 miles outside the corridor represents which of the following?

a A non-threatening target.

b A very threatening target.

c A target on the border between two threat levels.
d A target that is out of the possible range.

3 20 miles outside the corridor represents which of the following?
a A non-threatening target.
b A very threatening target.
c A target on the border between two threat levels.
(1 A target that is out of the possible range.

4 850 miles per hour represents which of the following?

A non—threatening target.

A very threatening target.

A target on the border between two threat levels.
A target that is out of the possible range.

Q06”

5 Class 10 radar type represents which of the following?

a A non-threatening target.
b A very threatening target.
c A target on the border between two threat levels.
(1 A target that is out of the possible range.
6 +20 degrees of angle represents which of the following?
a A non-threatening target.
b A very threatening target.
c A target on the border between two threat levels.
d A target that is out of the possible range.
7 +10 degrees of angle represents which of the following?
a A non-threatening target.
b A very threatening target.
c A target on the border between two threat levels.
(1 A target that is out of the possible range.

193

Recall that these are the nine target attributes:
Speed Alit Size Angle IFF Direct Corr.St. Radar Range

8 Who can measure speed?

a CAD

b AWAC

c Cruiser

d CAD and Cruiser

e AWAC and Cruiser
9 Who can measure altitude?

a CAD

b AWAC

c Cruiser

(1 CAD and Cruiser

e CAD and AWAC

10 Who can measure range?

CAD

AWAC

Cruiser

CAD and Cruiser
AWAC and Cruiser

(DO-00"”

11 Who can measure IFF?

a CAD

b AWAC

c Cruiser

d CAD and AWAC

e AWAC and Cruiser

12 Who can measure size?

a CAD

b AWAC

c Cruiser

(1 CAD and Cruiser

e AWAC and Cruiser

13 Which of the following combinations represent a more threatening target?
Slow and land radar targets.

b High flying and fast targets.

c Descending and inside the traffic corridor targets.

d

e

33

Outside the traffic corridor and low flying targets.
Close and ascending targets.

LIST OF REFERENCES

LIST OF REFERENCES

Andrisani, P. J., & Nestel, G. (1976). Intemal-extemal control as a contributor to
and outcome of work experience. W, 61, 30—34.

Armelius, K. (1979). Task predictability and performance as determinants of

confidence 1n multiple-cue judgments. WM,
20, 19-25.

Averill, J. R. (1973). Personal control over aversive stimuli and its relationship to

stress. P h ' ll ' , 80, 286-303.

Baird, I. S., & Thomas, H. (1985). Toward a contingency model of strategic risk
taking- W. 19. 230-243-

Bakeman, R., & Helmreich, R. (1975). Cohesiveness and performance: Covariation

and causality m an undersea environment. MW
Payehalegy 11 478-489.

,_’

Bateson, N. (1966). Familiarization, group discussion, and risk-taking.,[oama1_et
W.2,119-129.

Bazerman, M. H. (1982). Impact of personal control on performance. Is added
control always beneficial? W, 61, 472-479.

Bazerman, M. H. (1984). The relevance of Kahneman and Tversky’s concept of
fiaming to organizational behavior. LoaraaLeLMaaagemeat, 10, 333-343.

Bernoulli, D. (1738). Specimen theoriae novae de mensura sortis. Cemmeatagj
WW. 5. 175-192 Translated by

L. Sommer (1954) as Expositions of a new theory on the measurement of risk.

W. 22. 23-26.

Bird, A. M. (1977). Team structure and success as related to cohesiveness and

leadership- We» 1.03. 217-223.

194

195

Brehmer, B. & Hagafors, R. (1986). Use of experts in complex decision making.

A Paradigm for the study of staff work QrgarLZatienalBehardeLmHuman
Deeisienlmesses. 3.8. 181-195

Bumstein, E. (1969). An analysis of group decisions involving risk (”the risky
shift") Humanjelatiens. 22. 381-395.

Clark, R. D. 111. (1971). Group induced shift toward risk: A critical appraisal.
MW. 15. 251-270.

Collins. B E-. & GnetzerW. H (1964) W
W. New York. Wiley.

Curley, S. R, Yates, J. F. ,& Abrams, R. A. (1986). Psychological sources of

ambiguity avoidance WW
Eroeesses. 3.8. 230-256

Dailey, R. C. (1978). Relationship between locus of control, perceived group

cohesiveness, and satisfaction with coworkers. MW, 52,
311-316.

Davidson, 11. SUPPCS. P.. & Siegel, S. (1957). MW
appreaeh. Stanford, CA: Stanford University Press.

deCharms, R. (1968). - .. ' -
w. New York. Academic Press.

 

Dobbins, G. H. ,& Zaccaro, S. J. (1986). The effects of group cohesion and leader

behavior on subordinate satisfaction. WW, _1,
203-219.

Dorfman, P. W., & Stephan, W. G. (1984). The effects of group performance on
cognitions, satisfaction, and behavior: A process model. M

MW. 19. 173-192.

Dutton, J. E., & Jackson, S. E. (1987). Categorizing strategic issues: Links to
organizational action. W. 1.2. 76-90.

Dyer, J. L. (1984). Team research and team training: A state-of-the-art review. In

F. A. Muckler (Ed.) W, (pp. 285-323). Santa Monica,
CA: Human Factors Society.

Einhorn, H. J., & Hogarth, R. M. (1981). Behavioral decision theory: Processes of
judgment and choice. Annaal Review at Eeyehelegy, 32, 53-88.

196

Ellsberg, D. (1961). Risk, ambiguity, and the Savage axioms. W
fieenemies. 25. 643-669.

Ericsson, K. A., & Simon, H. A. (1980). Verbal reports as data. W
191152.81. 215-251.

Ericsson, K. A.. & Simon. H. A. (1984). WWW
Cambridge, MA: MIT Press.

Evans, C. R., & Dion, K. L. (1991). Group cohesion and performance: A meta-
analysis. Smallfirwmfieseareh. 22. 175-186.

Fagley, N. S., & Miller, P. M. (1987). The effects of decision framing on choice of

risky vs certain Options WWW
firms. 32. 264- 277

Fagley, N. S. & Miller, P. M. (1990). The effects of framing on choice:
Interaction with risk-taking propensity, cognitive style, and sex. 13152113111!

211W. 1.6. 496-510
Field, R. H. G. (1982). A test of the Vroom-Yetton normative model of leadership.

MW. 62. 523-532.

Field, R. H. G., & House, R. J. (1990). A test of the Vroom-Yetton model using

manager and subordinate reports. f li h l , 15, 362-
366.

Fleishman, E. A., & Zaccaro, S. J. (1992). Toward a taxonomic classification of
team performance functions. Initial considerations, subsequent evaluations and
current formulations. In R. W. Swezey & E. Salas (Eds. ) M

W New Jersey: ABLEX

Glass. D 0. &Singer. I E (1972) W
W. New York: Academic Press.

Greenberger, D. B., & Strasser, S. (1986). Development in application of a model of

personal control in organizations. W, 11, 164-
177.

Greenberger, D. B. ,Strasser, S., Cummings, L. L, & Dunham, R. B. (1989). The
impact of personal control on performance and satisfaction. W

W. .3. 29-51

197

Griffith, J. (1989). The Army’s new unit personnel replacement and its relationship
to unit cohesion and social support. WW, 1, 17-34.

Gross, K. (1901). W. New York: Appleton.

Hackman. J. R- (1990)- W- San FranciSCO.
CA: Jossey-Bass.

Hagen, O. (1979). Toward a positive theory of preferences under risk. In M. Allais

& 0. Hagen (Eds. ) W. (pp
271-302). Dordrecht, Holland. D. Reidel.

Hartnett, J. J ., & Barber, R. M. (1974). Fear of failure in group risk taking. British
Jeamal ef Seeial aae Clinig Peyehelegy, 13, 125-129.

Heilman, M. E., Homstein, H. A., Cage, J. H., & Herschlag, J. K. (1984).
Reactions to prescribed leader behavior as a function of role perspective: The

case of the Vroom-Yetton model. Wag, 69, 50-60.

Hoogstraten, J., & Vorst, H. C. M. (1978). Group cohesion, task performance, and
the experimenter expectancy effect. We, 11, 939-956.

Hollenbeck, J. R., Sego, D. J. Ilgen, D. R. ,Major, D. A. (1991.) W

 

h(Tech Rep No 91- -1) East
Lansing: Michigan State University, Departments of Management and

Psychology.

Ilgen, D. R., Major, D. A., Hollenbeck, J. R., Sego, D. J. (in press). Decision
making in teams: Raising and individual level decision model to the team

level. In R. Guzzo and E. Salas (Eds. )T earn 111ng reign mmng ia
W. San Francisco: Jossey Bass.

Jackson, S. E., & Dutton, J. E. (1988). Discerning threats and opportunities.
Warmly. as. 370387.

Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision
under risk. M, g, 263-291.

Kahneman, D., & Tversky, A. (1984). Choices, values, and frames. Ameg'm
Psychelegist, 32, 341-350.

198

Kameda, T., & Davis, J. H. (1990). The function of the reference point in individual

and groan risk decision making Qrganmtienekfiekaeramm
W525. .6. 55-76

Lamm, H, & Meyers, D. (1978). Group induced polarization of attitudes and

behaviors In L Berkowitz (Ed) Wired
253511619“, 11. New York: Academic Press.

Langer, E. J., & Roth, J. (1975). Heads I win, tails it’s chance: The illusion of
control as a function of the sequence of outcomes in a purely chance task.

MW. 32. 951-955.

Langer, E. J., & Rodin, J. (1976). The effects of choice and enhanced personal
responsibility. A field experiment in an institutional setting. learn a1 ef

fiemdinrandiaiakfiseekday, :L, 191-198.

Levin, I. P., Johnson, R. D., & Davis. M. L. (1987). How information frame
influences risky decisions: Between-subjects and witlrin-subjects comparisons.

IarrraLefflaariflseekelay. 8. 43-54.

Levinger, G., & Schneider, D. J. (1969). A test of the risk as a value hypothesis.
MW. 2. 165-169.

Libo, L. M. (1953). MW. Ann Arbor: University of

Michigan, Institute for Social Research.

Lichtenstein, S., Fischhoff, B., & Phillips, L. D. (1982). Calibration of
probabilities: The state of the art to 1980. In D. Kahneman, P. Slovic, & A.

Tversky (Eds.) Wares. (PP- 306-
334). Cambridge: Cambridge University Press.

Loke, W. H. (1989). The effects of framing and incomplete information on
judgments Inmdefifianomiefiseehdae. 19. 329- 341

MacCrimmon, K. R., & Larsson, S. (1979). Utility theory: Axioms versus

”paradoxes." In M. Allais & O. Hagen (Eds.) WW
theALlaiapmdea, (pp. 333-410). Dordrecht, Holland: D. Reidel.

Marquis, D. G. (1962). Individual responsibility and group decisions involving risk.
[negatijai Management Review, 3, 8- 23.

McClelland, D. C. (1961). W. New York: Van Nostrand.

McClelland, D. C. (1975). flwer; jifhe inner exmtiena. New York: Irvington.

 

199

McClelland, D. C., (1982). Leadership motive pattern and long-terrn success in
management WWW fl. 737-743

McClelland, D. C., & Winter, D. (1969). W. New
York: Free Press.

McCord, M., & de Neufville, R. (1983). Empirical demonstration the expected
utility decision analysis is not operational. In B. P. Stigum & F. Wenstop

(EdS)fio_u_dake.se__ut_1knLand_rrs__tkanauh_anp.rat_ens. (pp 181-200)
Dordrecht, Holland: D. Reidel.

McGrath. 1- E. (1984). W Englewood CliffS.
NJ: Prentice-Hall.

McGuire, T. W. ,Kiesler, S. & Siegel, J. (1987). Group and computer-mediated

discussion effects 1n risk decision making. W
fiseckdae 32. 917- 930

Meyers, D. ,& Iamm, H. (1975). The polarizing effect of group discussion.
AmerieaaSeiektist. 63, 297-303.

Miller, D. T., & Ross, M. (1975). Self-serving bias 1n the attribution of causality:
Fact or fiction? Esyehelegigifletin, 81, 213-225.

Miller, S. M. (1977). Controllability and human stress: Method, evidence, and
theory. Behavier Re_sear;eh and Therapy, 111, 287-304.

Morgan, B. B. Glickman, A. S. Woodward, E. A. ,Blaiwes, A. S. & Salas, E.

(1986). Measarement ef team mhmete 1a a Nayy eavimnmen t (Tech. Rep.
No. NTCS TR- -86-0.14) Orlando, FL. Navy Training Systems Center.

Mosteller, F., & Nogee, P. (1951). An experimental measurement of utility. 161mm
eLfieLik'edEeerteme. 32. 371-404

Mudrack, P. E. (1989). Defining group cohesiveness: A legacy of confusion? Semi
Gram. 2Q. 37-49-

Neale, M. A. ,,Bazerman M. H. ,,Northcraft G. B. ,&Alperson, C. (1986).
”Choice shift effects” in group decisions. A decision bias perspective.

lnternatrenakleurrrdakfimdfiremfieseareh. 2, 33-42.

Pepitone, A., & Kleiner, R. (1957). The effects of threat and frustration on group
cohesiveness. Ieumdakknarnarandmrdlseekdae. 3A. 192-19.

 

200
Platt, J. R. (1964). Strong inference. Seieage, 146, 347-353.

Pruitt, D. G. (1971). Choice shifts 1n group discussion. An introductory review.
WM. 29. 339- 360

Rothbaum, F., Weisz, J. R., & Snyder, S. S. (1982). Changing the world and
changing the self. A two-process model of perceived control. learn a1 ef

fierserrakteandjardfiseekdae 42 5-37

Rotter, J. B. ( 1966). Generalized expectancies for internal versus external control of

reinforcement. h l A 1i ,,8_Q 1-28.
Savage, L. J. (1954). MW. New York: Wiley.

Schoemaker. P. J. H. (1980). W
W. Boston: Martinus Nijhoff.

Schurr, P. H. (1987). Effects of gain and loss decision frames on risky purchase

11680113110“. W. 2.1.. 351-353-

Sitkin, S. B., & Pablo, A. L. (1992). Reconceputalizing the determinants of risk
behavior. MW. 11. 938-

Slovic, P., Fischhoff, B., & Lichtenstein, S. (1982). Facts versus fear:
Understanding perceived risk. In D. Kahneman, P. Slovic, & A. Tversky

(Eds.) W. (on- 463-489)-
Cambridge: Cambridge University Press.

Stasser, G., & Titus, W. (1985). Pooling of unshared information in group decision
making: Biased information sampling during discussion. iearnai ef

fiersaelflandioejdlseekdae. 38. 1467-1478.
Stoner, J. A. F. (1961). A mmpgisen ef ineivieeai gig green eggieiens invelviag

tisk. Unpublished master’s thesis, School of Industrial Management,
Massachusetts Institute of Technology.

Sutton, R. I. ,& Kahn, R. L. (1986). Prediction, understanding and control as
anecdotes to organizational stress. In J. Lorsch (Ed. ) Haadimekef
etgaaiaatioaauieaayiet, 272-285. Englewood Cliffs, NJ: Prentice-Hall.

Tetrick, L. E., & LaRocco, J. M. (1987). Understanding, prediction, and control as
moderators of the relationships between perceived stress, satisfaction, and

psychological well-being. ieemai ef Appliee Psyehelegy, 12, 538-543.

201

Thaler, R. H., & Johnson, E. J. (1981). Gambling with the house money and try to
break even: The effects of prior outcomes on risky choice. Maaagement

We 26: 643-660°

Thompson, S. C. (1981). Will it hurt less if I can control it? A complex answer to a

simple question. W, m, 89-101.

Tversky, A., & Kahneman, D. (1981). The framing of decisions and the psychology
of choice. Seigrt‘e A1, 453-458.

Van Bergen, A. , & Koekebakker, J. (1959). Group cohesiveness in laboratory
experiments. Aetafiseekdaiea. 16. 81-98.

Vinokur, A. (1971a). Cognitive and affective processes influencing risky taking 1n
groups= An espected utility approach Ionmalktfiesenaumknejaid
fiseekdae. 29. 472486

Vinokur, A. (1971b). Review and theoretical analysis of the effects of group
processes upon individual and group decisions involving risk. Psyelaelegieai
fidletin.16. 231-250.

Vlek, C., & Stallen, P. J. (1980). Rational and personal aspects of risk. Aeta
We ii. 273-3“)-

von Neumann, J., & Morgenstem, O. (1944). WW
@3119]:- Princeton, NJ: Princeton University Press.

von Neumann, J ., & Morgenstem, O. (1947). W
behavier, 2nd edition. Princeton, NJ: Princeton University Press.

Vroom, V. H, & Jago, A. G. (1978). On the validity of the Vroom-Yetton Model.
IonmdeLarqlfiseekdae .63. 151- 162.

Vroom. V. H.. & Yetton. P. W. (1973). Laderskrrundeedsrmndeng
Pittsburgh: University of Pittsburgh Press.

Wallach, M. A., Kogan, N., & Bem, D. J. (1964). Diffusion of responsibility and

level of risk taking In groups IdkndafikbnermekandfiareLfiseekdae.
68, 263-274.

Weinstein, N. D. (1980). Unrealistic optimism about future life events. LeewaLet
fiersenaliteanuaidfiseckdae. 32. 806-820

 

202

White, R. (1959). Motivation reconsidered: The concept of competence.
We fly 297-333-

Whyte. 6. (1989). GroUPthink reconsidered. WW. 1A.
40-56.

Yinon, Y., & Bizman, A. (1974). The nature of affective bonds and the degree of
personal responsibility as determinants of risk taking for ”self and others. "

fidletrn.ef_tke_fi5edrdromre_$aiete, A, 80-82.

Zaccaro, S. J ., & McCoy, M. C. (1988). The effects of task and interpersonal
cohesiveness on performance of a disjunctive group task. leuteaLeLAepjiee

Sedelfisedhdae. 18. 837-851.

Zuckerman, M. (1979). Attribution of success and failure revisited, or: The
motivational bias is alive and well in attribution theory. Mei

We .41: 245-287°

"111111111111111'11111