 

15..
M

g I
W"

...1

 

II i .2 I
stﬂuonlc at. u .

0 .,
2n... 3:52
: 6.. .. ,.......R»?.;

2: sir...-
. 57.1

 

 

$5; 1.4

 

THESIS

(7 7 f

z

lCHlGAN T TE U

Iu‘ir Iminil“mmlﬂilimumilil

3 3 01050 3237

 

 

 

 

This is to certify that the

dissertation entitled

EFFECT OF SIMULTANEOUS VERSUS SEQUENTIAL DISPLAY OF
VISUAL INFORMATION ON DECISION ACCURACY: MODERATING
EFFECTS OF DECISION CONTEXT

presented by

LINDA R. ELLIOTT

has been accepted towards fulﬁllment
of the requirements for

PHILOSOPHY degreein BUSINESS ADMINISTRATION

//o/ZZIQM—

Major professor

Date 11/22/96

MS U is an Affirmative Action/Equal Opportunity Institution 0-12771

 

 

 

LIBRARY

Michigan State ,
University

 

 

 

PLACE iN RETURN BOX to remove thin checkout from your record.
TO AVOID FINES roturn on or botoro date duo.

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU it An Atfirmotivo Action/Equal Opportunity Instituion
1

 

EFFECT OF SIMULTANEOUS VERSUS SEQUENTIAL DISPLAY OF VISUAL INFORMATION
ON DECISION ACCURACY: MODERATING EFFECTS OF DECISION CONTEXT

By

Linda R. Elliott

A DISSERTATION

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

DOCTOR OF PHILOSOPHY

Department of Management

1996

ABSTRACT

EFFECT OF SIMULTANEOUS VERSUS SEQUENTIAL DISPLAY OF VISUAL INFORMATION
ON DECISION ACCURACY: MODERATING EFFECTS OF DECISION CONTEXT

By

Linda R. Elliott

This effort examined three avenues of research (i.e. decisionmakng; automatic versus eﬂ‘ortful
cognitive processes; visual cue characteristics) to predict patterns of decision error when complex
information is visually displayed. Complex information is increasingly represented by perceptual
cues conﬁgured to enhance an intuitive recognition-based response, through simultaneous display of
visual cues conﬁgured as an overall pattern. These conﬁgural displays have been associated with
better performance on divided attention tasks, when compared to cues that are usually perceived
separately. However, it is proposed here that sequential presentation of cues that are part of a
conﬁgural pattern will be more easily perceived as separate and will facilitate more effortful
consideration of cue information and will result in higher accuracy when the decision context is
favorable. F avorability of context is determined by both task characteristics (degree of time
pressure, ambiguity of information, conﬂict among cues, irrelevant frame information) and
individual characteristics (cognitive ability, cognitive style). In addition, while research participants
in the sequential condition were predicted to perform more accurately than those in the simultaneous
display condition when the decision context is favorable, they were expected to be less accurate
when the decision context is degraded. Nine perceptual cues such as location, size, color, length and
direction of arrows, and audio pitch/tempo were chosen to be easily interpreted, in this case to
represent location, size, type of radar (red = hostile), speed and direction, and electronic signal.
Results supported expectations when comparing performance under task conditions that were most

favorable versus most degraded. Information display condition differed on type and degree of

decision error. Subjects in the simultaneous condition were more susceptible to error due to
averaging cues inappropriately. In contrast, subjects in the sequential condition were more
susceptible to anchoring-and-adjustment error, where preliminary information is weighted more than
subsequent information. It was also found that subjects in the simultaneous condition made their

judgements more quickly, even when they had more time to make their decision.

ACKNOWLEDGMENTS

I could not have accomplished this task without the support and guidance of many individuals. I had
the encouragement and support of family and friends to help me through the entire process. Thanks
Momli Also, special thanks go to Teri Mercatante and Tom Watson, who can always be counted on
to urge me to follow my heart. I owe much to the guidance, leadership, and advocacy of Colonel
Ronald C. Hill. My motivation was further reinforced by MSU faculty, who inspired interest in
every topic. As for the dissertation experience, the content of the dissertation is the least of what is
learned. I can only express deep thanks to all committee members, with special thanks to my chair,
John Hollenbeck, for his encouragement, guidance, and adherence to high standards. Finally I would
like to thank my current “boss”, Samuel G. Schiﬂett, for his steady expressions of conﬁdence and
character. All should strive to meet his standards of spirituality, compassion, and dedication to

scientiﬁc advancement and operational problem solving.

iii

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................... vii
LIST OF FIGURES ............................................................................................................ viii
CHAPTER 1: INTRODUCTION .......................................................................................... 1
Individual limitations and biases .................................................................................. 1
Investigation of visual displays .................................................................................... 3
Purpose of the Present Study .................................................................................................... 6
Background ............................................................................................................................. 7
Visual Cue Characteristics and Task Performance ....................................................... 8
Perception of visual cues ................................................................................. 9
Type of visual display and performance ......................................................... 10
Simultaneous Versus Sequential Displays .................................................................. 12
Resistance to Degraded Circumstances .......................................................... 13
CHAPTER 2: VISUAL DISPLAYS, PERCEPTION, AND DELIBERATION ..................... 15
Display Characteristics And Decision Process ............................................................ 16
Cognitive continuum theory ........................................................................... 17
Analytical Versus Intuitive Processes ......................................................................... 19
General deﬁnitions ........................................................................................ 20
Intuition as an automatic process ................................................................... 22
Two dimensions to classify intuitive processes ................................................ 24
Fitting diverse descriptions into the framework ............................................... 26
Recognition versus deliberation ..................................................................... 30
CHAPTER 3: PREDICTIONS ............................................................................................. 32
Effect Of Information Display ................................................................................... 32
Simultaneous display of visual information ................................................... 32
Sequential display of visual information ........................................................ 33
Interactions between display and decision context .......................................... 35
Task Characteristics Affecting F avorability: Main Effects .......................................... 36
Time pressure ............................................................................................... 36
Information ambiguity ................................................................................... 37
Information conﬂict ....................................................................................... 39
Irrelevant ﬁame information .......................................................................... 39
Task Characteristics: Interaction Effects .................................................................... 39
Display and time pressure ............................................................................. 40
Display and ambiguity .................................................................................. 41

iv

Display and information conﬂict ................................................................... 43

Display and frame ......................................................................................... 44
Individual Characteristics: Main Effects ..................................................................... 46
Cognitive ability ............................................................................................ 46
Cognitive style .............................................................................................. 46
Individual Characteristics: Interaction Effects ............................................................. 47
Display and cognitive ability ......................................................................... 47
Display and cognitive style ............................................................................ 48
Summary ................................................................................................................... 49
CHAPTER 4: METHOD ...................................................................................................... 51
Sample ...................................................................................................................... 51
Task .......................................................................................................................... 5 1
Information cues ........................................................................................... 52
Decision alternatives ..................................................................................... 57

Cue Interpretation Rules ............................................................................................ 58
Ambiguous decision events ........................................................................... 60
Feedback ...................................................................................................... 61
Research Design ........................................................................................................ 62
Experimental Manipulations ...................................................................................... 63
Sequencing of cue presentation ...................................................................... 63
Decision frame: previous error ...................................................................... 63

Time pressure ............................................................................................... 64
Information ambiguity .................................................................................. 65
Information conﬂict ....................................................................................... 65
Construction of decision events ..................................................................... 65
Procedure .................................................................................................................. 66
Research participant training ......................................................................... 67
Hypotheses and Data Analysis ................................................................................... 68
Hypotheses: Task characteristics ................................................................................ 68
H1 - Time pressure ........................................................................................ 68

H2 - Information ambiguity ......................................................................... 68

H3 - Information conﬂict ............................................................................. 68

H4 Frame ................................................................................................... 69
Hypotheses: Interactions between display and task characteristics ............................... 69
H5 - Time pressure and display .................................................................... 69

H6 - Information ambiguity and display ......................................................... 69

H7 - Information conﬂict and display .......................................................... 70

H8 - Decision frame and display ................................................................... 70
Hypotheses: Individual differences .............................................................................. 70
H9 - Cognitive ability .................................................................................. 71

H10 - Cognitive style .................................................................................... 71
Hypotheses: Interactions between display and individual differences .......................... 71
H1 1 - Display and cognitive ability .............................................................. 71

H12 - Display and cognitive style ................................................................. 71

CHAPTER 5: RESULTS ..................................................................................................... 72

Overview .................................................................................................................. 72
Task Characteristics .................................................................................................. 73
Time pressure ............................................................................................... 73

Time pressure and display ............................................................................... 75
Ambiguity ..................................................................................................... 77
Ambiguity and display ..................................................................................... 78

Cue conﬂict .................................................................................................. 80
Conﬂict and display ........................................................................................ 81

Frame .......................................................................................................... 82

Frame and diisplay ....................................................................................... 84
Individual Characteristics .......................................................................................... 86
Cognitive ability ........................................................................................... 86
Cognitive ability and display ......................................................................... 86
Cognitive style .............................................................................................. 88
Cogitive style and display ............................................................................. 89
Exploratory Analyses ................................................................................................ 90
Manipulation of favorability ......................................................................... 92
Deceptive targets .......................................................................................... 94

Order effects ................................................................................................. 99
Performance over time .................................................................................. 101
CHAPTER 6: DISCUSSION .............................................................................................. 105
CHAPTER 7: REFERENCES ............................................................................................. 127

vi

LIST OF TABLES

Table l - Task Characteristics Associated With Intuitive and Analytical Responses ............... 18
Table 2 - Framework For Conceptualizations of Intuition ..................................................... 25
Table 3 - Attributes and Ranges Underlying Perceptual Cues ............................................... 55
Table 4 - The Seven Decision Alternatives ............................................................................ 57
Table 5 - Rules of Engagement ............................................................................................. 59
Table 6 - Underlying Quantitative Model .............................................................................. 60
Table 7 - Scoring of Accuracy ............................................................................................. 62
Table 8 - Generation of Targets: Within-Subjects Variables ................................................. 66
Table 9 - Between-subjects Variables ................................................................................... 72
Table 10 - Impact of Time Pressure on Decision Accuracy .................................................... 73
Table 11 - Impact of Display and Time Pressure on Decision Accuracy ................................ 75
Table 12 - Impact of Ambiguity on Decision Accuracy .......................................................... 77
Table 13 - Impact of Display and Ambiguity on Decision Accuracy ....................................... 78
Table 14 - Impact of Cue Conﬂict on Decision Accuracy ......................................................... 80
Table 15 - Impact of Display and Cue Conﬂict on Decision Accuracy ...................................... 81
Table 16 - Impact of Decision Frame on Decision Accuracy .................................................. 83
Table 17 - Impact of Decision Frame on Response Mean ....................................................... 84
Table 18 - Impact of Display and Decision Frame on Response Mean .................................. 83
Table 19 - Descriptives for SAT and ACT Scores .................................................................. 86
Table 20 - Impact of Display and Cognitive Ability on Decision Accuracy ............................. 87
Table 21 - Descriptives for MBTI Sensing and Intuitive Scales .............................................. 89
Table 22 - Impact of Display and Cognitive Style on Decision Accuracy ................................ 89
Table 23 - Impact of Display on Decision Accuracy .............................................................. 91
Table 24 - Impact of Favorability and Display on Decision Accuracy .................................. 93
Table 25 - Impact of Deception on Decision Accuracy ............................................................. 95
Table 26 - Impact of Time Pressure, Display and Deceptiveness on Decision Accuracy .......... 96
Table 27 - Impact of Ambiguity, Display and Deceptiveness on Decision Accuracy ............... 98
Table 28 - ANOVA Results for Order and Display ................................................................ 99
Table 29 - Regression Weights for Decision Responses by Display: Comparison to Ideal
Regression Weights and Algorithm Weights .................................................. 101
Table 30 - Main effect of Display on Later Performance ....................................................... 102
Table 31 - Impact of Favorability and Display on Later Performance ..................................... 103

vii

LIST OF FIGURES

Figure l - Types of Visual Cues .......................................................................................... 10
Figure 2 - Overall Model ..................................................................................................... 36
Figure 3 - Predicted Effect of Time Pressure ........................................................................ 40
Figure 4 - Predicted Effect of Ambiguity .............................................................................. 42
Figure 5 - Predicted Effect of Conﬂict .................................................................................. 43
Figure 6 - Predicted Effect of Irrelevant Frame Information .................................................. 45
Figure 7 - Predicted Effect of Cognitive Ability .................................................................... 45
Figure 8 - Predicted Effect of Cognitive Style ...................................................................... 45
Figure 9 - Overall Model ..................................................................................................... 49
Figure 10 - Graphic Display of Target With All Nonthreatening Cues .................................... 53
Figure 11 - Graphic Display of Targets With All Threatening Cues ........................................ 54
Figure 12 - Threat Values Associated With Each Cue ............................................................. 56
Figure 13 - Expected Interaction Between Visual Display Condition and Favorability

of Decision Context .......................................................................................... 110

viii

Chapter 1
INTRODUCTION

For the past 40 years, researchers used rational models of decisionmaking to describe and
predict decision making behavior (Edwards, 1954; Einhom & Hogarth, 1981; Hammond, 1955;
Payne, Bettman, & Johnson, 1988; Savage, 1954). These models serve as criteria by which
decisionmaking performance can be assessed when the decision task is essentially rational.
Certainly, criticisms have been raised when the rational model is applied to decision situations
which include nonrational and/or suboptimal elements. Such elements include variables such as
values, commitments, personal impact, societal factors, and overarching goals (Beach & Lipshitz,
1993; Hemstein, 1990; Orasunu & Connolly, 1993; Thibaut & Walker, 1978; Zey, 1992);
dynamic complexity (Beach et al., 1993; Cohen, 1993); tasks involving problem structuring and
interpretation of ambiguous cues (Mintzberg, Raisinghani, & Theoret, 1976; Orasunu et al.,
1993), social decision making (Fiske & Taylor, 1991), and organizational decision making (March
& Shapira, 1992; Simon, 1955; 1990; 1992), among others (Jacob, Gaultney & Salvendy, 1986;
Payne, Bettrnan, & Johnson, 1992; 1988). Nontheless, when the decision task is composed of
quantitative elements which can be calculated to produce an unequivocally correct decision, the
rational model serves as providing an “ideal” from which decision errors can be identiﬁed and
described.

Individ_u_al limitations and bias_e_s. Within this boundary condition of rational tasks (i.e.

tasks in which decision mics can be applied to ascertain a correct response), there is no doubt that

systematic patterns of decision error occur (Edwards, 1954; Jacob, et al., 1986; Kahneman &

Tversky, 1972;1(ahneman, 1991;1(ahneman, Slovic, & Tversky, 1982; Massaro, & Cowan,
1993;Meeh1, 1957; Payne, et al., 1988; Simon, 1955; Slovic, Fischoff, & Lichtenstein, 1977;
Stevenson, Buscmeyer, &Naylor, 1990; Tversky, 1972, 1977). These errors are usually attributed
to limitations in human information processing capabilities such as working memory, long-term
memory, and processing speed. The predictability of these errors suggests that these constraints
result in simpliﬁcations, biases and distortions in individual information processing and
decisionmaking, resulting in patterns of error such as overconﬁdence, representativeness, ﬂaming
effects, availability, and illusory correlation (Kahneman & Tversky, 1972; 1982; Kahneman,
Slovic & Tversky, 1982; Kleinmuntz, 1990).

Decision errors have also been associated with socio-cognitive and emotional factors. Thus
decision errors can result from cognitive limitations such as a limited working memory capacity
(Massaro & Cowan, 1993), and also from biases in social perception (F iske & Taylor, 1991;
Nisbett & Ross, 1980), differences in attitudes or values (Hammond, Harvey, & Hastie, 1992)
and/or emotional distress (Parkinson & Manstead, 1992).

From these observations, alternative models of decisionmaking have arisen to better describe
and predict individual decisionmaking. In order to explain decisionmaking in dynamic, complex
situations, alternatives to the rational paradigm have been proposed that are more intuitive in
nature (Hammond, Hamm, Grassia, & Pearson, 1987; Klein 1993; Klein, Orasunu, Calderwood,
& Zsambok, 1993; Orasunu & Connolly, 1993; Zey, 1992). These alternatives focus on
explaining decision processes and the role of cognitive ability and expertise, under conditions that
are typically complex, ambiguous, and/or stressful.

Alternatives to the rational paradigm are generally simpliﬁed models of decisionmaking, to
better reﬂect individual decision processes under conditions more representative of naturally

occurring conditions of complexity, ambiguity, uncertainty, and time pressure. Many alternatives

to rational models exist, such as image theory (Beach, 1990; 1993; Beach & Lipshitz, 1993; Beach
& Mitchell, 1978), recognition-primed decisionmaking (Klein, 1993), satisﬁcing (Simon, 1955),
cognitive continuum theory (Hammond, et al., 1993), and various simpliﬁcation strategies
(Stevenson, Buscmeyer, & Nayior, 1990). In support of the efﬁcacy of these simplifying and/or
satisﬁcing strategies, Dawes (1982) demonstrated that simplifying heuristics such as averaging cue
weights can still provide accurate estimates. The particular weighting scheme had less impact on
accuracy; the deletion of one or more cues had a greater effect in producing error. Similarly,
Gilliland, Schmitt, and Wood (1993), Kerstholt (1992), and Payne, Bettman, & Johnson (1988)
demonstrate tradeoﬁ‘s of different decision models and accuracy. Hogarth (1981) argues that
biases assumed to be dysfunctional can actually be functional when decisions are made in a
naturalistic and dynamic setting over a continuous length of time. It can be seen that alternative
models have arisen in response to decision outcomes not well explained by traditional models of
rational deliberation and from ﬁndings that alternative strategies can be effective in some
circumstances. It is evident that we need to more fully develop and specify alternative models that
ﬁt these decision situations.

Investigation of visual displays. At the same time that human limitations and individual
diﬂerences were investigated as sources of decision error, a complementary stream of research
focused on effects of characteristics of the decision task that elicit systematic errors in
decisionnnaking performance. Thus characteristics such as infomnation complexity, volume,
tempo, ambiguity, and differences in how the information is presented are studied as
manipulations. This perspective drives the ﬁeld of cognitive engineering, where artiﬁcial
intelligence, decision aiding, and information presentation displays are designed in order to create a
decision context which maximizes advantages of computer driven analysis of data versus reliance

on human cognitive processes.

The manner in which decisionmakers are presented with information has been related to
systematic differences in decisionmaking performance. The results regarding relationships among
visual cue characteristics and performance suggest performance can be enhanced by matching
these characteristics with the cognitive demand of the task. Visual display characteristics have
been related to decision processes and performance (Andre & Wickens, 1992; Bennett & Flach,
1992; Boles & Wickens, 1987; Coury, Boulette, & Smith, 1989; Coury, & Boulette, 1992;
Hammond et al., 1987; Sanderson, Flach, Buttigieg, & Casey, 1989; Wickens, 1986; 1990;
Wickens & Andre, 1990). Visual cues have been found to affect to differentially affect
performance, depending on whether the decision was to be based on focused attention, divided
attention, or the integration of all cues. In general, it has been found that cues which are easily
distinguished results in higher accuracy in tasks requiring focused attention, and cues conﬁgured as
an overall pattern are better when cues must be integrated (Bennett & F lach, 1992; Pomerantz &
Pristach, 1989; Sanderson, et al., 1989).

Theories relating irnfonnation display characteristics to decision performance are mainly
based on relationships among basic cognitive processes such as attention and working memory and
performance on visual search and discrimination tasks. Triesman (1986) provides a review of the
literature on ﬁndings related to object perception and performance. Decision performance is
enhanced when the information is presented in such a way as to reduce cognitive effort in data
collection, working memory, and integration of cue information.

Visual representation of infomnation usually accomplishes integration of information by
capitalizing on the human capability of pattern recognition. In their discussion of graphic displays

and cognitive processes, Bennett and Flach (1992) state:

“There appears to be a clear consensus that performance can be improved by providing
displays that allow the observer to utilize the more efﬁcient processes of perception and pattern
recognition instead of requiring the observer to utilize the cognitively intensive processes of
memory, integration, and inference. Thus this type of display (geometric object formats) has the
potential to improve decision-making performance by shifting the burden of responsibility from
cognitive processes that are severely limited (e.g., working memory) to cognitive processes that,
with learning, are virtually unlimited (e.g., object perception and pattern recognition)” (p. 514).

Studies of visual display characteristics support the use of cues conﬁgured as a pattern when
cue attention must be divided or when information must be integrated. However, in this study, an
argument will be presented that separable cues can be more effective, and that the favorability of
the decision context is an important moderating variable. Simultaneous presentation of visual cues
has been associated with a more intuitive decision process (Hammond, et al., 1987). While
descriptions of intuition vary, intuitive decision processes in general have been characterized as
less effortful, less rational, and more prone to error. Thus, there appear to be conﬂicting evidence
regarding the efﬁcacy of visual pattern displays and the more intuitive decision process which has
been associated with it.

Visual displays are extensive in aviation and military settings, such as air trafﬁc control,
aircraft cockpits, and command-and-control centers. They are also common in business, academic,
and recreational settings. As distributed units with unique perspectives (e.g. military theater of
war, global corporations, multi-national scientiﬁc research) become more extensively linked,
coordination will be even more dependent on visual representation of strategic information. Given
the widespread prevalence of visual display of information, questions naturally arise regarding the
impact of task characteristics and individual characteristics on decision performance within this

context.

Purpose of the Present Study

This study examines the impact of individual and task characteristics on decision accuracy in
a rational task where infomnation is presented in a visual display format. It is predicted that
sequential versus simultaneous display of visual information would affect the degree of cognitive
effort in the decisionmaking process, and would be typiﬁed by different patterns of decision error.
In addition, it is further predicted that the patterns of error would be affected by the favorability of
the decision context, as indicated by characteristics of the decisionmaker and of the decision task.
Given an objective decision task which requires integration of several information cues, the
sequential display of visual cues is predicted to result in higher performance than simultaneous
displays when conditions are favorable. Sequential display of visual infomnation is expected to
direct attention to each cue and the decision rules associated with cue interpretation, resulting in
more effortful processing of information.

Sequential consideration of each one is expected to lead to more accurate assessments when
the process is allowed to occur under favorable conditions, that is, when there is complete and
certain information, and sufficient time to consider all information and decision rules. Favorable
conditions also include characteristics of the decisionmaker; that is, decisionmakers who are highly
capable for the task. Capability can be a function of skill, expertise, cognitive ability, and/or
cognitive style.

Ifdecisionmakers base their decisions on an overall impression, as predicted to occur with
simultaneous presentation of a pattern of visual information, the impression can usually be
modeled by an averaging or “summing up” strategyuwhen cues are highly intercorrelated, this
stratgey would be very effective. Even when cues are not usually intercorrelated, the averaging
strategy woud be effective when all cues are in agreement. However, if complex decision rules

must be considered, such that the value of one cue cannot be interepreted without knowing the

value of one or more other cues, an averaging strategy can be quite ineffective. Of course, if
decisionmakers have memorized all possible patterns, the averaging strategy would be minimized-
the decisionmakers would respond with certain recognition. When the number of possible patterns
is small, recognition is easily achieved; however, when there are numerous pattems, it is more
diﬁcult.

When the decision context degrades, research participants in the sequential display condition
are expected to be more vulnerable to error than those in the simultaneous display condition. The
higher amount of cognitive effort expected to occur in this condition would also be more likely to
be disrupted when conditions are unfavorable. Thus, characteristics such as time pressure,
infomnation ambiguity, and information conﬂict were expected to interfere with sequential and
rule-based consideration of cues. In contrast, research participants in the simultaneous display
condition are expected to react quickly with a more automatic response, less vulnerable to time
pressure or considerations regarding missing information.

In this study favorability of decision context is indicated by four task characteristics and two
individual difference variables. Task characteristics expected to affect favorability include time
pressure, information ambiguity, information conﬂict, and irrelevant frame information. A
favorable task context occurs when there is low time pressure, no ambiguity, no conﬂict, and no
irrelevant information. Favorability of the decision maker will be measured by cognitive ability
and cognitive style. A desirable decisionmaker in this context is one who has high cognitive ability

and a deliberative, fact-based style of information processing.

Background
It is predicted that simultaneous presentation of visual cues will be perceived as a holistic

pattern and will facilitate intuitive, recognition-based assessment. In research regarding visual

information display, the simultaneous display created for this study is consistent with the concept
of a conﬁgural visual display. A conﬁgural display has emergent properties such that an overall
pattern can be perceived.

In contrast, sequential presentation of the same cues is expected to focus attention on
individual cues and the decision rules by which these cues interact to determine the correct
assessment. Sequential presentation is therefore expected to enhance decision accuracy when this
more deliberate and effortan process can occur, with complete information, explicit decision rules,
and no interuptions or limitations to the infomation processing required to apply the decision rules.
When the decision context is degraded, deliberation is hampered, interupted, or stopped;
consequently decision performance is degraded. These effects are expected to be less degading for
decisionmaking in the simultaneous condition. The following section reviews existing knowledge
related to information display characteristics and task performance.

Visual Cue Characteristics and Task Performance

As decision tasks become more cognitively demanding, individual differences in abilities such
as processing speed, working memory, and dual task capability deﬁne the limits of decision
performance. At the same time there has been corresponding effort to analyze decision tasks for
their cognitive demands, to (a) predict performance given a particular task and (b) redesign the task
to facilitate performance.

Many studies have been performed which demonstrated that decision performance is affected
by the manner in which information is presented to research participants ( Barnett & Wickens,
1988; Boulette, Coury & Bezar, 1987; Carswell & Wickens, 1987 1990; Coury & Pietras, 1989;
Woods, Wise, & Hanes, 1981). These ﬁndings have been explained by relating display

characteristics to cognitive aspects of the decision task.

Perception of visual cues. Bennett and Flach (1992) reviewed ﬁndings regarding the
congruence of visual display characteristics and the cognitive demands of the decision task.
Several theories relate visual display characteristics to types of cognitive task (Barnett & Wickens,
1986; Casey, 1986; Pomerantz & Pristach, 1989; Sanderson, Flach, Buttgieg, & Casey, 1989;
Wickens, 1986), predicting higher performance when the visual display characteristics are
congruent with the cognitive demand of the decision task.

The visual representations of information are often distirnguished on the basis of the
separability of individual information cues. Separable cues are deﬁned by a lack of interaction
among the stimuli, such as color versus shape. With these cues, the cognitive demand is
characterized by case of selective attention, somewhat more effort for divided attention, and no
gain ﬁom redundancy. In contrast, integral cues are redundant, such that a change in one cue
results in a change in the other perceptual cue, such as a trafﬁc light having “stop” represented by
both color (red) and location (top). This results in a redundancy gairn, but makes it difﬁcult to focus
attention on one cue only, thus making selective attention and divided attention more eﬂ‘ortful.

A third category, display oﬂconﬁgural cues, is a mix of separable and integral characteristics.
With a conﬁgural display, each cue can be perceived as a separate entity, but new emergent
properties are created when these cues are perceived together (See Figure 1). For example, if the
length of ﬁve individual lines were separate cues conﬁgured in a geometric shape, the overall shape
is an emergent property. Changes in the shape of the overall ﬁgure can be more easily perceived
than changes in individual line lengths. Bar graphs can also contain an emergent feature, if there is
attention to general trends in the graph (e.g. attending to whether the bars increase or decrease in a
particular pattern). Changes in the emergent property may better elicit perceptions of change as
opposed to attending to the individual cues separately. Figure 1 provides examples of each type of

visual cue and the cognitive processes which are best accommodated by each type of cue. The next

section discusses the matching of visual cue characteristic and cognitive task demand in further

detail.

Figpre 1. Types of Visual Cues.

/\

 

 

 

 

 

 

33.3
<——o
24 /
Separable Cues Integral Cues Configural Cues
(Control Task) (Redundant Task) (Divided Attention Tgk)
(Selective Atteption Tag) (Correlated Task) (Integr_ation Task)

 

Type of visual display and performance. The effectiveness of these three types of cue display
has been related to the type of cognitive task being performed. Bennett and Flach (1992) describe
four types of cognitive tasks which should be considered when designing displays. They differ in
the focus of attention which is required.

In the ppm task, the individual need only attend to one cue which varies, and the other cues
are held constant. For this type of task, displays of separable cues are considered most conducive
to effective performance. In the selective gttention task, the individual still focuses on one cue, but
the other cue(s) varies. In this type of task, separable cues are also most appropriate. In the

correlated or redund_a_n_t_ task, cues vary simultaneously and discrimination can be made by

11

attending to either or all cues. For this task type, integral cues would provide a redundancy gain.
In the divided attention t§s_k or cue integration task, variations in all cues must be considered, and
Bennett suggests that a conﬁgural display, with its separable elements and emerging qualities,
would facilitate performance (Bennett et al., 1992), based on the cumulative ﬁndings of many
studies.

Consistent with Bennett’s conclusions, Wickens’ compatibility of proximity theory predicts
effects of visual displays based on perceptual aspects of divided and focused attention (W ickens,
1986; Wickens & Andre, 1990). He states that if the task requires focused attention on a single
cue, the display with separable cues would be more effective, but if the task requires integration, an
object (conﬁgural) display is better. For example, Wickens and Andre (1990) found that when
indicators of aircraft stall danger were distinct in color focused attention was improved but
integration of information was disrupted. If the cues were presented as an object rather than a bar
graph, infomnation integration was improved but focused attention performance degraded.

Research supports the ﬁnding that when the task is one that requires cue integration, the use
of a conﬁguration of cues (object) is effective. Coury et al. (1989) found that such displays are
particularly effective when the decision task is based on multiple cues where the values of these
cues are correlated. In these circumstances object displays have been found to be consistently
superior to alphanumeric displays. This is explained as the result of enabling the subject to
recognize a unique object conﬁguration for a particular decision response category. The physical
representation of these cues creates a conﬁguration with unique features which can be mapped to
the underlying state represented by the cues. According to Wickens (1986) object displays enable
rapid holistic integral processing of system cues.

Sanderson et a1. (1989) adds that the display does not need to be an object per se, such as a

geometric shape, but any ﬁgure can be conﬁgural if it contains an emergent feature. For example,

12

if one uses a conﬁguration of three lines formed as a triangle, if the lines represent values which
must be monitored, when the lines change in length, so do the angles of the triangle. In this way
the angles of the triangle are emergent features which represent additional or consolidated
infomnation. Sanderson points out that even bar graphs (separable cues) can have emergent
features. To illustrate, the viewer may attend to a conﬁguration of the bars (increasing to the right
or left) rather than the values of the bars per se. According to Gamer (1978) emergent properties
may be functions of symmetry, repetition, intersections, conjoining, and angular separation. The
attention to the emergent features is enabled by the pattern recognition capability of the viewer
(Bennett & F lach, 1992; Sanderson et a1, 1989).

§imultaneops versus Seguential Display of Visual Cugs

The simultaneous presentation of visual/audio cue information in this study provides a
conﬁgural display with strong emergent features. The visual representations were chosen to be
easily interpreted-size is indicated by size; location by location; speed and direction by length and
direction of arrows, etc. (see Figure 10). In the sequential display the same cues are presented
beginning with one cue (location) and adding cues one at a time such that when all cues are
presented the display is equivalent to the simultaneous condition.

I expect that sequential presentation of the same information produces more separable
information cues, because they are presented one at a time in a cumulative fashion. Thus, in
contrast to conclusions that conﬁgural displays are best for divided attention tasks (Bennett et al.,
1992; Wickens & Andre, 1990), it is argued here that sequential presentation will lead to greater
accuracy when decision rules are complex and conditions for decisionmaking are favorable.

The task in this study requires consideration of all cues, according to interactive decision
rules. Because the cues are not redundant, and because simply averaging the information cues can

result in error, in this situation the more conﬁgural (simultaneous) display does not have a strong

13

advantage. The emergent feature in the simultaneous condition is the overall impression one
perceives from the pattern of cue information. However, an appraisal based on an overall
impression is expected to be associated with increased error compared to assessment based on
systematically considering each cue. Sequential presentation of cue information is expected to
facilitate more effortful and systematic consideration of cues and decision rules.

This task has nine infomnation cues and four decision rules, and thus systematic and effortﬁil
consideration of each cue will result in more accurate assessments. When there are many cues and
cue interactions to consider, a large variety of possible patterns can result, making it diﬁicult to
easily recognize each urnique pattern. Thus reliance on an overall impression by an inexperienced
decision maker is likely to result in increased error. In contrast, the sequential display of these cues
should facilitate more effortful consideration of each one and decision rule, and thereby lead to
more accurate assessments.

Woe to degraded circumstan_ces_. While the sequential display of visual information is
expected to result in higher decision accuracy, it is also predicted that more deliberate
consideration of cues will be more vulrnerable to degrading task conditions such as time pressure,
ambiguity, and conﬂict and to degraded information processing due to differences in individual
cognitive style or cognitive ability. When conditions are degraded, the sequential display condition
is expected to result in lower accuracy due to greater interference. In contrast, the simultaneous
display condition is expected to be more robust to degrading conditions.

The rationale for this expectation may be better understood within the context of the cognitive
task demand of the decision. The decision task used in this study is complex, such that many
different visual patterns can arise, making recognition more difﬁcult. In addition, interactions
among the cues must be considered in order to calculate the correct assessment. The task demands

analytical deliberation of the cue information in order to calculate the correct assessment.

14

It is expected that sequential presentation of cue information will be associated with a higher
degree of effortful deliberation on the part of the decision maker, and this match with the task
demand will facilitate performance. On the other hand, it is expected that simultaneous
presentation of cue information will be associated with a higher degree of intuitive, recognition-
based decisionmaking. This recognitional process is expected to be less accurate than effortful
deliberation of complex infomnation, but more resistant to detrimental effects. Thus, while
research participants in the sequential condition are expected to perform more accurately under
favorable conditions, research participants in the simultaneous display condition are expected to
perform more accurately under degraded conditions, due to the type of decision error elicited by

these displays. Decision processes and associated decision errors are discussed in chapter 2.

Chapter 2

VISUAL DISPLAYS, PERCEPTION, AND DELIBERATION

In this study, it is expected that simultaneous versus sequential visual display of information
will be associated with differences in amount and type of decision error. These errors are thought
to arise from elicitation of decision processes that differ in degree of deliberate versus automatic
processing of information. Sequential presentation of visual cues are expected to elicit a more
controlled cognitive process, while simultaneous presentation is expected to be perceived as a
holistic pattern, initiating a recognition response which varies in level of certainty. When
recognition is certain, the response is fast and accurate. When recognition is uncertain, the
response is more intuitve and less accurate.

Effortful delibertion of information cues and decision rules should result in higher accuracy
than intuitive processing when information cues are certain and decision rules are explicit.
Sequential presentation of cue information is expected to result in higher accuracy than the
simultaneous display condition. However, effortful processing requires more cognitive resources
and thus requires accurate information, an explicit algorithm, and enough time to process all
information according to rules. Thus, it is more vulnerable to conditions such as time pressure,
uncertainty, and conﬂicting infomation. A recognition-based response is expected to be more
resistant to degrading contextual factors such as time pressure. The following section reviews the
current tlninking regarding the conceptualization of these different processes, and the impact of

infomation display characteristics on decision processes.

15

16

Display Characteristics And Decision Process

As noted previously, researchers investigating effects of visual displays of information expect
to enhance decision accuracy through design of displays which allow ease in perception and pattern
recognition (Bennett & F lach, 1992). Similarly, in this study, it is proposed that the automatic
and/or intuitive processing of information elicited by visual display features have different patterns
of associated errors when compared to more effortful deliberation of information. Further,
aspectes of the decision context are expected to moderate the impact of visual display
characteristics and decision performance.

The proposal that visual display will affect type of error is not new. Hammond (1987)
included display features as one of the task characteristics affecting the manner in which
information is processed. He predicted that congruence between the process (intuitive or
analytical) required by the underlying cognitive demand of the decision task and the process elicited
by surface characteristics of the decision task, such as infomnation display features, would
facilitate decision performance. Similarly, in this study the cognitive demand is expected to be
associated with differing requirements for etfortful and deliberate analysis of information and that
task characteristics are expected to affect the level of cognitive effort in the decisionmaking
process. However, speciﬁc mechanisms and predictions differ from Hammond’s.

This study distirnguishes between display conditions and other characteristics of the decision
context. Cognitive demand and display of information are expected to be of fundamental
importance in inﬂuencing the process by which the decisionmaker responds. Other characteristics,
such as time pressure, ambiguity, conﬂict, and capability of the decision maker, formulate the
decision context. The favorability of the decision context is distinguished from display condition
and are instead proposed as important moderator variables affecting the relationship between

display and performance.

17

While these decision context characteristics are included in Hammond’s list of decision task
characteristics affecting decision process, they are distinct from the cognitive demand of the task.
The underlying cognitive task demand can be represented by the complexity and speciﬁcity of
decision mics used to assess and integrate information cues. Thus, if a decision task had explicit
and complex mics by which information should be considered, the cognitive demand indicates that
systematic deliberation is congment with the task. Other characteristics are proposed to be
characteristics of the decision task context. These variables may include many aspects, such as
time pressure, incomplete information, conﬂicting information, consequences of error, and
requirements for sustained performance over a long period of time; or as characteristics of the
decisionmaker, such as expertise, experience, fatigue, cognitive ability, and cognitive style.

Congmencc between cognitive task demand and cognitive process is expected to be an
important predictor of decision accuracy, as predicted by Hammond (1987), and it is predicted here
that effortful and analytical processing of information in a task comprised of complex and explicit
decision mics will result in higher accuracy, but only when the decision context is favorable.
When the decision context in unfavorable, attempts at effortful processing is impeded and
interrupted, resulting in lower accuracy compared to a more intuitive, less effortful approach to
decisionmaking.

First, Hammond’s cognitive continuum theory will be described, followed by a section
discussing the existence and differentiation of effortful analysis versus more intuitive cognitive
processes. Chapter 3 describes and discusses speciﬁc predictions made in this study.

Qggnitive continuum theory. Hammond’s cognitive continuum theory maintains that human
information processing varies along a continuum that ranges from very analytical to very intuitive.
Witlnin the middle of this continuum lies “quasirationality,” which he states is similar to “common

sense” or “bounded rationality” (Hammond, 1987). His cognitive continuum theory is based on the

18

assumption that characteristics of the decision task will elicit analytical or intuitive decision
processes, which in turn are associated with different types of decision error. He related task
characteristics such as type of information (perceptual or quantitative), time pressure, uncertainty,
etc. (see Table l) to elicitation of analytical versus intuitive decision processes.

According to Hammond, if many or all of the task characteristics are those that elicit one
form of decisionmaking, that type of decision process will be elicited. For example, he states that
if the task includes (a) many redundant cues, (b) the cue values are continuous, (c) the cues are
measured perceptually, and (d) there is no underlying principle for assessing the cues, the subject
will respond with intuition, with the corresponding decision weighting scheme characterized by a

simple summation of the cues.

Table 1.

Task Characteristics Assﬂnted with Intuitive and Analyticgl Responses (Hammond et al., 1987)

 

 

Intuition Analysis
1. Number of cues > 5 < 5
2. Measurement of cues perceptual objective
3. Distribution of cue values Continuous Unknown
4. Redundancy among cues High Low
5. Decomposition of task Low High
6. Degree of certainty Low High
7. Cue-criterion relation Linear Nonlinear
8. Weighting of cues Equal Unequal
9. Organizing principle Unavailable Available
10. Display of cues Simultaneous Sequential
11. Time period Brief Long

 

The characteristics of a decision task were combined to form a task continuum index,
described as the extent to which a decision task will elicit analytical to intuitive responses. A

cognitive continuum index was generated to represent the degree to which a decision response was

19

analytical versus intuitive. Congmencc between the task continuum index and the cognitive
continuum index was investigated for impact on decision performance. Hammond reported that
task characteristics did elicit predicted variations in intuitive and analytical decision strategies.

While Hammond reported a direct comparison of analytical versus intuitive response, there is
little knowledge as to the relative importance of these task characteristics in this elicitation, and
alternative explanations can be generated for the results. For example, his manipulation crossed
three clusters of task display thought to elicit intuition (visual pictures under high time pressure),
quasi-rationality (bar graphs under time pressure), and analysis (requirement to produce formulas
to derive and justify responses, no time pressure) with tlnrec types of task content corresponding
with intuition (highway aesthetics), quasirationality (highway safety), and analysis (lnighway
capacity). Research participants were experts with these tasks.

Contrary to predictions, performance was often higher when the display was intuitive (i.e.
visual picture) even when the decision content was analytical (assessment of highway capacity).
This may also be by reason of familiarity and experience in that research participants are used to
using visual pictures for these determinations, and have already associated visual representations of
highways to aspects such as safety and capacity. While results were not entirely as predicted, they
demonstrate the interesting aspect of this study, that visual display elicited an intuitive response as
measured by Hammond, and that the visual display manipulation resulted in higher performance

even when the task was analytical.

Analy_tical Versus Intuitive Processes

Predictions from display theories assume that intuitive decision processes can be elicited and
differentiated ﬁ'om more analytical processes. While Hammond’s cognitive continuum dreary has
intuitive decision process as a primary constmct, the constmct is not speciﬁcally described. At

this time the constmct of “intuition” is almost useless due to the lack of a consistent deﬁnition.

20

Existing deﬁnitions were reviewed in order to identify inconsistencies and clarify the constmct as
used in this study. Because inconsistencies were found, a framework was generated to identify the
differentiating characteristics of the intuitive decision process.

This section will discuss the existence of intuitive versus analytical processes, with the
intention of organizing the various descriptions of intuition, and specifying what is meant by
intuition in the context of a rational task. Many researchers have discussed distinctions between an
effortful, deliberate, analytical decision process versus one that is nonanalytical, or intuitive.
While most agree on what is meant by an analytical process, there are various and somewhat
contradictory deﬁnitions of intuitive processes. Intuition at this time is a very fuzzy notion.

Genergl deﬁnitions. What Hammond (1987) refers to as an “analytical” process, others have
referred to as “rational” in nature. When compared to the nonanalytical “intuitive” process, a
primary distinction is the cognitive effort involved in the process. This distinction can be captured
by the use of the term “deliberative” process, which is descriptive of the cﬂ‘ortful deliberation
inherent in a rational or analytical task. In this discussion, the process described as the eﬁ‘ortful,
analytical, sequential processing of infomnation will be referred to as the “deliberative” process.

Intuition has been described as the counterpoint to rational tlnought (Epstein, 1994). For
example, Beach and Mitchell (1978, Mitchell & Beach, 1990) refer to "non analytical" strategies,
where deliberate consideration of all alternatives is bypassed. Similarly, Kahneman and Tversky
(1982a; 1982b) describe intuitive judgnents as composed of an unstmctured process which
includes no deliberate calculation or analytical metlnod. However, it is not very precise to deﬁne
the intuitive process by what it is not.

Most researchers will agree that intuition is not deliberate, effortfnnl, or analytical. There is
growing agreement that unconscious, implicit processing of information does occur (Reber, 1993;

Schactcr, 1987; Seger, 1994). However, this still does not provide us with a succinct deﬁnition of

21

intuition. Deﬁnitions of intuition cover many different phenomena, ranging from scientiﬁc insight
to parapsychic prophesy (Agor, 1989; Vaughan, 1989). The following section reviews existing
formulations of intuition, followed by the conceptualization adopted for this study.

Some authors have equated intuition as any human decision process that is not a speciﬁc
decision algorithm (Denes-Raj & Epstein, 1994; Kleinmuntz, 1990). For example, Epstein (1994)
proposed two fundamental processes, the rational system and the experiential system, as means by
which individuals assess the environment and make judgnents. According to Epstein, the
experiential system includes heuristics, intuition, and affect. Epstein represented intuition as
holistic, affective, associationistic, and more automatic than the separate process of objective
reasoning. The intuitive process was also represented as the source of all deviations from
rationality. In contrast, the rational system was presented as an information processing capability
that is conscious, analytic, and logical, with more highly differentiated responses.

Many other researchers have proposed multiple information-processing systems that include a
process which is unconscious, automatic, and/or affective in nature. These proposals range from
the historical contributions of philosophers such as Aristotle (experiential versus rational
krnowledge) to conceptualizations from psychodynamic, experimental-cognitive, developmental,
and social-cognitive perspectives (Epstein, 1994). Epstein provided a compelling network of
arguments that these variations on multi-modal processes converge on the identiﬁcation of a
conscious, rational, deliberative system and a second system that has characteristics of
automaticity, experiential knowledge, and emotionality.

It is proposed here that the intuitive decision process associated with conﬁgural visual
displays is based on the cognitive process of recognition. This recognition process is expected to
differ from deliberation by the manner in which cue infonnnation is processed. Recognition is based

on simultaneous processing of one information such that the decision object is perceived as a

22

whole, without conscious deliberation of each cue. Research reported within the ﬁeld of cognitive
science supports this view.

Intuition as an automgtic process. Certain recognition is an automatic process. Logan (1988)
provides a clear discussion of automatic processes, and while there are intriguing discussions as to
the explanatory mechanisms underlying automaticity, there is agreement that automaticity is fast,
effortless, consistent, and unavailable to conscious awareness. Processes described as automatic
include procedural knowledge (Anderson, 1983), implicit or tacit knowledge, and implicit learning
(Reber, 1993). These processes are considered to be automatic, entailing little conscious eﬂ’ort.

Cognitive scientists have proposed a model of information processing, based on cognitive
architectures that enable simultaneous processing of information (Lord & Maher, 1991). Lord and
Maher provide an extensive review of cognitive theory in decisionmaking. In their review they
outline the differences between symbolic and connectivist cognitive architectures in describing
information processing, and relate intuition to expert problem solving and parallel processing.
They state "implicit knowledge such as intuition may be more aptly explained by connectivist
architectures."

Lord and Maher also stated that problem solving by individuals with expertise appears
consistent with the parallel processing described witlnin connectivist architectures (Kosslyn &
Anderson, 1992). The lack of articulation commonly associated with experts attempting to explain
their decisions or procedures may be due to the mismatch in using a symbolic process, that is,
articulation, to describe an essentially connectivist process (Lord & Maher, 1991).

Several similarities can be drawn between the intuitive/deliberative distinction and the
concepts of controlled versus automatic processing of irnfonnnation (Shiffrin & Shneider, 1977;
Schneider & Shiffrin, 1977). According to their theory, automatic processing results from

activation of a learned event, without conscious effort, and without reducing attentional capacity.

23

In contrast, controlled processing requires attention, is capacity-limited, usually serial in nature,
and is controlled by the subject. This theory is focused on attention and search mechanisms,
particularly search of familiar targets among distracters. The authors demonstrated differences
and dynamics in the elicitation and performance of automatic detection versus controlled search
processes. In the same way, the intuitive process has the fundamental features of automatic
processing, that is, of low subject awareness and rapid parallel processing without deliberation.
The deliberative process shares the same features as controlled processing-«he subject actively and
serially processes the information. In this proposal the concepts of automatic versus controlled
attention and search mechanisms are extended to more complex cognitive functions of assessment,
judgnent, and decisionmaking.

In their review of current literature regarding information processing, Massaro and Cowan
(1993) distinguished between fully serial information processing (only one item at a time),
capacity-limited parallel processing (a limited number of items at a time), and capacity-ﬂee
parallel processing. They drew an analogy of infomnation processing capacity as a bridge, the
width of which (a) determines how many items can "cross" at a time and (b) is affected by the
automaticity of the process. Massoro also portrayed infomnation processing theory as including
"continuous representations similar to activation in many connectionist models." (p. 387)

Because recognition is often explicit and intentional, one may argue that it does not ﬁt the
criteria used for an automatic process (i.e. unintentional, involuntary, effortless, autonomous,
occurring outside of awareness). However, as Bargh (1989) noted, there are automatic processes
that have one but not all of the characteristics. Using his categorization, recognition processes can
be explicit and intentional (goal-directed), yet still be an essentially automatic process. Ashby and
Maddox (1990) discussed various theories of the recognition mechanism, stating that general

recognition theory is one based on "experienced categorization." They add, "Although it may be at

24

times difficult to determine which decision region an exemplar is irn, the theory predicts that once
this is established, categorization is essentially automatic" (p. 601.).

Some have described intuition as an implicit process (Lord & Maher, 1991), but one can
distinguish between implicit processes versus recognition-based judgnents. An implicit process by
deﬁnition refers to acquisition of knowledge without explicit awareness (Seger, 1994). Similarly,
Reber (1993) states, "Implicit learrning is the acquisition of knowledge that takes place largely
independently of conscious attempts to learn and largely in the absence of explicit knowledge of
what is acquired." In contrast, according to Simon, recognition is explicit (Cooper, Schacter,
Ballesteros, & Moore, 1992; Simon, 1989; 1986).

This overview of cognitive research reveals great progress in deﬁning cognitive processes that
are automatic and based on simultaneous processing of infomnation. It has even been suggested
(Reber, 1993) that automatic, unconscious cognitive processes are the default condition. Rather
than proving an intuitive process was utilized, Reber challenges that the burden of proof should be
to prove that conscious deliberation has taken place.

Two dimensions to clpssify intuitive processes. The diversity of constructs or perspectives on
intuition appears to differ on two basic dimensions: the degree of certainty of the response, and the
degree to which information or knowledge is consciously held, ranging from explicit awareness of
facts (prinning recognition-based responses) to implicit/subjective knowledge characterized by gut
feelings or spiritual experiences (see Table 2).

Some deﬁnitions of intuition draw upon or are consistent with the notion of subjectivity and
implicit processes, in that he subject is not aware of the rationale of the decision, it was based on a
“feeling”, the rationale cannot be articulated, etc. This perspective can be contrasted with

deﬁnitions of intuition based on recognitional processes that arise from expertise, explicit training,

25

which can be associated with a high degree of certainty even if the subject did not process the
information in a systematic, effortful manner.
Table 2

_F'_r_amework for Diverse Conceptualizations of In_tuition

 

 

 

 

Awareness Level of certainty
Low High

Implicit “Feeling of knowing” “TranscendencclBelief”
(Rowan, 1989) (Rowan, 1989)
(Barnard, 1938)

Explicit “Best Guess” “Insight/Aha”
(Barnard, 193 8) (Goldberg, 1989)
(Hammond, 1993) (Klein, 1993)
(Simon, 1992) (Simon, 1992)

 

When information regarding cues and decision mics is certain and explicit, intuition is that of
automatic or semi-automatic cognition (as opposed to affect or parapsychic phenomena).
Recognition can be conceptualized as ranging in certainty, from an amorphous "feeling of
knowing" to "particular certainty." Of course, some researchers (Allwood & Montgomery, 1987
as an example) would not agree, as they distinguish intuition from recognition, stating that intuition
is a function that is inarticulate, a simple “feeling of knowing” that occurs somewhere between a
random guess and certain recognition. Their deﬁnition of intuition is limited to uncertain events,
certain recognition would not be considered intuitive, even though it is automatic. It is apparent
that intuition is not a precise term; rather, it is commonly used to refer to diverse characteristics.
The two dimensions of cxplicitrness and certainty appear to capture basic differences among the

descriptions of intuition.

26

Fitting diverse descriptions of intuition into the frpmework. Researchers with different views
on intuitive processes can be placed within Table 2, according to descriptions of level of conscious
awareness of infomnation/rationale, and the level of certainty of the response.

Many researchers agree upon the theme of intuition as rapid and unconscious cognition.
Rowan (1989) described intuition as "knowledge gained without rational thought" (p. 84). In
ﬁirther describing intuition, he stated:

Elusive as it is, we do know certain characteristics of this inner impression or hunch. It

concerns relationships, involves simultaneous perception of a whole system, and can draw a

conclusion—not necessarily correctuwithout procwding through logical intermediary steps.

That's why intuition comes with that queasy feeling of almost but not quite knowing. (p 85)

According to Rowan, intuition is perceived as arising from subconscious processing of
information. The process itself was described as a spontaneous ﬂash of insight which cannot be
articulated by the decision maker. This lack of articulation is a common theme in deﬁnitions of
intuition.

Rowan's conceptualization of intuition is more in alignment with intuition arising from
implicit processing. This description of intuition as spontaneous, inarticulate insight is consistent
with the notion of innsight arising from processing of implicit cues. Other researchers, discussed
below, will describe intuition with characteristics more consistent with explicit cues

Barnard (1938) provided a description of intuition consistent with that of Rowan's deﬁnition
of intuition as spontaneous insight, Bamard distinguishes between what he called "logical" and
"nonlogical" processes:

By "logical processes" I mean conscious thinking which could be expressed in words or by
other symbols, that is, reasoning. By "non logical processes" I mean those not capable of

being expressed in words or as reasoning, which are only made krnown by a judgnent,

27

decision, or action. This may be because the processes are unconscious, or because they
are so complex and so rapid, often approaching the instantaneous, that they could not be
analyzed by the person witlnin whose brain they take place. The sources of these non-
logical processes lie in physiological conditions or factors, or in the physical and social
environment, mostly impressed upon us unconsciously or without conscious eﬁ’ort on our
part. They also consist of the mass of facts, patterns, concepts, techniques, abstractions,
and generally what we call formal knowledge or beliefs, which are impressed upon our
minds more or less by conscious effort and study. (p.302).

Bamard's conceptualization of intuition also includes characteristics of rapid, unconscious
processing. It is interesting that he speciﬁcally distinguished and included both implicit and
explicit cues as eliciting intuitive processes. He also distinguished between implicit and explicit
processes as resulting in intuition when he stated that cues "impressed upon our minds more or less
by conscious effort" can also result in intuition. Thus, deliberate learning of explicit krnowledge is
described as a source of intuitive phenomena.

Goldberg, (1989) described intuition as comprised of deliberate as well as spontaneous
components. Factual knowledge and logical analysis are considered along with more non rational
inputs such as feelings, dreams, hunches, and spontaneous insight. Thus a scientiﬁc discovery may
be realized in an apparently instantaneous patterning of krnowledge already attained by the
individual. Goldberg describes how a number of scientists attribute their discoveries to
spontaneous insight, and how these "sudden leaps to understanding" relate to processes of
discovery, creativity, evaluation, operation, prediction, and illunnination. This representation of
intuition can be said to rely more on explicit cues (krnowledge already attained by the individual)

resulting in an intuitive insight with a high degree of certainty.

 

 

28

Hammond (1989) proposed several factors as differentiating intuitive decisionmaking, such as
low cognitive control, rapid infomnation processing, low conscious awareness, and low conﬁdence
in the answer (low level of certainty). There is a common theme that intuitive decision processes
can difficult to articulate when decisionmakers are asked to describe their decision process. Thus,
one may have a "gut feeling" that a decision is correct, but be unable to systematically describe the
basis for the decision. Hammond's deﬁnition of intuition also states they are typiﬁed by decisions
rapidly made at a low level of explicit awareness.

Hammond’s conceptualization supports the notion of intuition as rapid and automatic. His

assertion that intuition is marked by low conﬁdence, however, may not be the case for all instances.

The framework provided here allows intuitive processes to lead to certain responses. Low
conﬁdence in response typiﬁed by lack of articulation is consistent with intuition derived from
subjective / implicit cues.

Klein and his associates (Klein, 1993; Klein, Calderwood, & Clinton-Cirocco, 1988; Klein,
Orasunu, Calderwood, & Zsambok, 1993) presented a model of intuitive decisionmaking based on
recognition that also describes intuition as rapid and automatic. He has reported several studies
focused on decisionmaking of experts under time pressure. From his investigations of military
oﬁ'ncers in combat and ground ﬁre crews, Klein has observed that complex decisionmaking under
time pressure does not follow a deliberate process. Experts faced with stressful decisions with
signiﬁcant consequences under high time pressure do not report having generated more than one
option at a time. The situation is "recognized", a solution is brought to mind, and that solution is
utilized or rejected without generating alternative solutions. Witlnin Klein's recognition-primed
decisionmaking, expert decisionmaking under conditions of time pressure relies on intuitive
assessments, based on the recognition of a scenario, rather than a deliberate sequential assessment

of relevant factors.

29

Klein's conceptualization of intuition is more in alignment with explicit cues. His focus was
on individuals explicitly trained to a high degree of expertise. The result is rapid and automatic
decisionmaking that appears intuitive, yet arises from explicit cues and conscious processing of
infomation. The decisionmaking behavior of these experts under time pressure is rapid and
automatic, yet the underlying rationale for each decision could be articulated afterwards.

In his discussion of intuition, Simon (1992) succinctly states, "Intuition is nothing more and
nothing less than recognition." While we may be unaware of the process of recognition, we are
conscious of the fact that recognition took place. According to Simon (1990) recognition processes
underlie cognitive processes such as grand master chess playing, medical diagnosis, and reading.
Simon does not differentiate between intuition and rational thought:

“When the problems to be solved are more than trivial, the recognition processes have to be

organized in a coherent way and they must be supplied with reasoning capabilities that allow

inferences to be drawn from the information retrieved, and numerous chunks of information
combined. Hence intuition is not a process that operates independently of analysis; rather, the
two processes are essential complementary components of effective decisionmaking systems

(p.33). It is a fallacy to contrast "arnalytic" and "intuitive" styles of management. Intuition

and judgment are simply analyses frozen into habit and into the capacity for rapid response

through recognition" (p.38).

While Simon may not distinguish between intuitive and rational decision processes, one can
argue that "recognition" is a process more instantaneous and less conscious than deliberative
analysis, and is consistent with the category offered here regarding explicit cues. Simon
distinguishes between intuition as recognition and insight, deﬁned as spontaneous understanding on

a deeper level than recognition. This distinction may be difﬁcult to maintain. Witlnin the

30

framework, Simon's deﬁnition of insight would also ﬁt as an intuitive process resulting ﬁom
simultaneous processing of explicit cues leading to a high level of certainty.

Recognition versus delibcrmp. If intuition is based on recognition, the intuitive decision
process is quite different from a deliberative approach. In contrasting the deliberative process with
recognitiorn, two distinguishing characteristics are apparent. First and primary, recognition is fast
and relatively effortless. As stated earlier, recognition is a process that can occur instantly, when
the cue is deeply familiar. In contrast, the deliberative process requires time and effort to process
each cue, apply decision mics, generate possibilities and probabilities, evaluate each and identiﬁy
the correct or optimal decision. The second characteristic distinguishing recognition from
deliberation is that recognition is based on a holistic perception. Deliberation, in contrast, is by
deﬁnition a sequential process. This often involves consideration of each information cue,
application of decision mels, and generation of possible solutions and outcomes, performed in a
step-by-step fashion. On the other hand recognition usually occurs in an instantaneous fashion,
and is thus likely based on parallel, rather than serial information processing, to enable
simultaneous processing of many cues.

In summary, visual display characteristics have been differentially associated with
performance on different types of cognitive tasks. When the task requires integration, conﬁgural
represeentation has been recommended. However, conﬁgural displays have been associated with
intuitive decisionmaking. The intuitive process elicited by these conﬁgural displays is argued to be
a type of recognition-based response, which is faster and less effortful, but less accurate than
systematic deliberation of all information, when decision mics are complex.

Intuitive processes have been associated with an averaging error; that is, performance can be

modeled by a unit-weighted linear model. While averaging strategies can be robust, they can also

31

lead to large errors when the decision mics are interactive and the information cues are not
correlated.

When the decision task has complex mles by which information must be processed, more
effortfui and systematic analysis should lead to more accurate assessments than will an intuitive
process. However, effortful deliberation is expected to be more vulnerable to degradation by
unfavorable context characteristics, such as time pressure or low capability of the individual.
Thus, decision context is offered as an important moderator variable in modeling the effect of
visual display on decision process and patterns of decision error. These predictions are described

in detail in chapter three.

Chapter 3

PREDICTIONS

Effect 91’ Information Display

The preceding section discussed the various conceptualizations of intuitive processes in order
to clarify the intuitive constmct that is often stated to be elicited by conﬁgural displays of visual
information. The conclusion is that the intuitive response that is referred to is based on explicit
recognition processes that can vary in degree of certainty.

Simultaneous (My of visugl cues. Using the framework provided by Hammond (1989) the
primary eliciting factor for an intuitive recognition-based response is proposed to be the
simultaneous presentation of perceptual information, that is, presentation of a "pattern".
Simultaneous presentation of multiple perceptual cues is expected to induce a more intuitive
attempt to recognize the overall pattern in the display.

Intuitive decision making is thus presented as an automatic, recognition-based decision
process (Simon, 1992). It is this ability to recognize complex patterns which is most human, and
most difﬁcult for linear computer progranns to emulate. This pattern recognition process would
result rrom deliberate leaming processes; however, once the cues become familiar, the judgnent
process transitions from one of deliberation to that is more automatic and recognition-based.

It is expected in this study that simultaneous display of visual cues will facilitate an intuitive,
pattern recognition response, varying in degree of certainty. Even though the decision task requires
deliberation of interactions, it is expected that the simultaneous display of visual information easily

perceived as a holistic unit will induce a more intuitive, pattem-recognition response. If the

32

33

recognition response is uncertain, the decision maker may tinen consider cue information more
deliberately, to verify the overall impression, if time pemnits. Even so, the assessment would likely
be less systematic than when cues are deliberately considered to formulate a judgnent. Instead, the
cues would be scanned for consistency with the overall impression, and the impression adjusted
accordingly.

Seguential display of visual infomnation. The underlying cognitive task demands systematic
deliberation of information. In this study, the task was constructed such that infomnation cues
must be considered separately and systematically, in accordance to four decision mics. For this
type of task, it is expected that sequential display of visual information will in general result in
higher accuracy than simultaneous displays.

It is expected that serial presentation of cues would induce a deliberate, sequential approach
to information processing. This is counter to some previous research (Matin & Boff, 1988) which
suggested that serial presentation of information is processed more rapidly and automatically.
However, in their study, the stimuli consisted of single digits, presented serially (same location on
screen) or all at once, and it is a recall task. Research participants had to recall the infomnation in
order, and this would be easier when serially presented than when the information had to be
processed in a spatial nnanner (the ﬁrst digit is in the upper left window, the second in the upper
right window, ctc.). Thus their task differed in (a) nature of the task (recall versus
decisionmaking), (b) nature of the stimulus (quantitative versus perceptual) and (c) number of cues
(3 versus 9). While their conclusion was that sequential presentation of infomation is more
automatic, the difference can be attributed to differences in task demands. The task used in this
stnrdy provides the simultaneous display of infomnation with strong emergent features, that is, the
image can easily be comprehended as a whole. In addition, the previous study was a recall task,

whereas this task will require more effortful considerations of separable but interacting cues.

34

Sequential presentation of these perceptual cues is expected to facilitate consideration of
individual cues and their interactions. Research participants are told to consider each cue as it
related to its interacting cue. If one cue is safe the interaction is safe regardless of the threat level
of the other cue; therefore each tinreatening cue must be considered in light of the other interacting
cue. When cues are presented sequentially, in order of interacting cues, the subject will more
easily be able to interpret each speciﬁc interaction.

If the cues are presented sequentially, the "pattern" cannot be ascertained until all cues are
presented. Also, when the cues are presented singly, it is expected that the decision maker will
consider each cue as it is being presented. Hammond (1990) has reported results consistent with
these expectations. According to Hammond deliberative cognition will result if a decision task
presents nonredundant cues in a sequential fashion, along with an explicit principle, scientiﬁc
theory, or metlnod for organizing cues into a judgnent.

Additional support for the expectation that sequential cues would elicit a deliberative response
is supported by research regarding integral versus separable cues. It has been demonstrated that
displays with separable cues require the subject to serially process the information, and the
additional mental workload results in greater time and effort to process the cues (Coury &
Boulette, 1992). Sequential presentation of cues should also have the same effect as separable
cues. While a perceptual pattern is built, and can be seen at the very end, I expect that sequential
presentation of cues minimizes the perception of the whole, and focuses attention to the separate
cues.

Thus, two factors serve to elicit deliberation in this task. One is the underlying demand that
research participants consider cue interactions; a cue that is in itself threatening may not be
correctly assessed as tlnreatening if the interacting cue is safe. In addition, sequential display is

expected to result in a more deliberative response. Serial presentation is expected to induce serial

35

processing of cue infomnation and reduce parallel processing of the pattern of cue conﬁgurations.
Simultaneous presentation of multiple channels of cue information is expected to be more readily
perceived as a holistic pattern leading to recognitional responses. Presentation of cues based on a
variety of perceptual stimulinlocation, color, size, sound, etc., is expected to facilitate parallel

processing and recognition of the pattern of cue conﬁguration.

 

Interactions between displgLand decision conga. The major proposition of this dissertation
is that decisionmaking performance will depend on the interaction between display type and
characteristics of the task and the decision maker. It is proposed here that intuition as a holistic
recognition process can be effective, even in a decision task with objective cues and algoritluns for
accuracy, depending on the favorability of decision context. When conditions are ideal, the
sequential analysis of infomation is expected to be associated with greater accuracy. However,
when situational conditions are degraded, due to factors such as time pressure, conﬂicting cues,
and/or low cognitive ability, tinen sequential presentation of infomnation, with its requirement to
systematically deliberate the information, is expected to be associated with a higher rate of error.
In contrast, simultaneous presentation of perceptual information is expected to be less affected by
degraded conditions. The nature of this interaction is such that these conditions can result in
greater accuracy resulting from recognition-based responses. Figure 2 provides the overall model

of the predictions made inn this study.

Figpre 2. Overall Model

36

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

TASK
CHARACTERISTICS
INFORMATION DECISION
DISPLAY ACCURACY
4b
INDIVIDUAL
CHARACTERISTICS
o A \a
a c \‘A
7 3 \- +350
3 r +SIM
i a
0 C
n Y
Favorrable Unfaniorabie
FAVORABLE UNFAVORABLE
Low time pressure High time pressure
Certain irnfonnation Ambiguous information
No conﬂict Conﬂicting cues
High cognitive ability Low cognitive ability
Analytical cognitive style Intuitive cognitive style

Task Chgractgristics Affecting Favorability: Main Effect;

Figure 2 includes main effects of task characteristics on performance across both conditions.

The main effects for these characteristics are expected to be as follows.

Time pressure. Time pressure is known to have a negative effect on performance, due to

constraints on attentional and processing resources (Coury & Bouiett, 1992). According to

37

Hogarth (1975) the time required for decisionmaking is a function of task complexity. Other
characteristics inﬂuencing the effect of time pressure include extent of attentional resources,
mapping of stimuli to resources, task pacing, and type of display (Coury & Boulette, 1992).

In this study time pressure will range from having sufﬁcient time to complete the task to high
time pressure, where research participants report perceptions of time pressure. As time pressure is
increased, the cognitive demand of the task is inncreased, as decision makers will have to process
information more rapidly. Time pressure is expected to degrade accuracy by interrupting and/or
not allowing sufﬁcient processing of information.

In addition, the need to make an immediate judgnent will increase vulnerability to biases such
as central tendency (judgnents are less varied in assessment). Directional bias can also occur, if
the decisiomnaker has a preference for reducing error due to overly safe assessments (increasing
chance that the aircraft is really threatening and consequently takes hostile action) or to overly
aggressive assessments (increasing chance that the aircraft is really nontinreatening and is wrongly
attacked). Time pressure is expected to degrade accuracy in the simultaneous condition as well,
but to a lessor degree compared to the sequential display.

Information ambigpity It is reasonable to expect that information ambiguity, as deﬁned by
missing information, would be detrimental to performance. Decision makers will have decision
situations where (a) no information is missing, (b) some information is missing, and (c) nearly a
third of the information is nnissing. As information becomes more ambiguous, the decision maker
will not be able to calculate the correct outcome with certainty, even if there is no time pressure.

A review of decision performance under uncertainty emphasizes the role of ambiguity as a
source of error. Individuals are usually ambiguity-averse (Curley, Yates, & Abrams, 1986) and
have been willing to pay to avoid ambiguity (Kahn & Sarin, 1988). They assess lower

probabilities of occurrence for a single event, such as the probability of rain, but higher

38

probabilities for an outcome based on a series of probabilistic events (Boiney, 1993; Gettys et al
1982). Even statisticians have been found to over-rely on small numbers (T versky & Kahneman,
1982). Intuitive prediction of uncertain events relies almost exclusively on information particular
to the event, rather than information regarding the population (Bar-Hillel, 1982; Kahneman &
Tversky, 1982; Tversky & Kalnneman, 1982).

Tversky and Kahneman (1982) list numerous errors in human judgnent under uncertainty,
such as anchoring and adjustment, availability, and representativeness. While there has been a
great deal of research regarding decisionmaking under ambiguity or risk, most of it has been for
choice decisions involving probabilities such as gambling decisions (Kahneman et al, 1982). In
those situations, individuals are likely to violate SEU predictions by choosing the alternative with a
lower payoff that is more certain. Models have been offered to accommodate this phenomenon of
ambiguity aversion (Einhom & Hogarth, 1985; Kahn & Sarin, 1988) within the paradign of SEU
theory. However, they do not provide direct implications for the type and degree of decision error
that can arise from unavoidable ambiguity witlnin an objective task, beyond that of increased error.
F risch et al (1988) did provide a prediction similar to "regression to the mean" in that, given
extreme but limited information, the decision maker will assume the missing information will
provide a less extreme judgnent, if the decision maker cannot delay the decision.

A common assumption is that ambiguity leads to intuitive judgnent (Kahnennan et a1, 1982
Maharn, 1992; 1994). Discussion of error under ambiguity is therefore taken to be a discussion of
intuitive error. Tversky and Kahneman use this deﬁnition of intuition, stating "The reliance on
heuristics and the prevalence of biases are not restricted to laymen. Experienced researchers are
also prone to the same biases - when they think intuitively" (p. 18). That may be so if one deﬁnes

intuition as any human judgnent process which is not favorable. The overall ﬁnding here is that

39

ambiguity has been related to higher error in decisionmaking, and is thus expected to degrade
accuracy in general.

Information conﬂict. Conﬂict among information cues is expected to be detrimental to
performance in general, as it also creates ambiguity during information assessment. Conﬂict is
deﬁned as a decision situation where some of the cues indicate one level of assessment, and other
cues indicate a very different level of assessment. In the sequential condition, the conﬂicting cues
will be presented such that preliminary cue infomnation will conﬂict with subsequent cues. In the
simultaneous condition, the decision maker will also see the same conﬂicting cues, ratlner than
seeing a more easily recognized pattern with consistent cues. Conﬂict among cues is expected to
lower decision accuracy.

Irrelevant frame infomnagtion. Irrelevant background information is also expected to have a
detrimental effect on decision performance in general. It has been demonstrated that research
participants will respond differently to a decision problem depending on the manner in which the
problem or alterrnatives are phrased (Frisch, 1993), such as when the decision outcomes are worded
as a gain or a loss (Kahneman & Tversky, 1982). In this study background information which
describes a previous error (with disastrous consequences) is expected to have a frame effect, where
research participants will be more inclined to avoid the error previously described, and be more
likely to make errors in the opposite direction.

Task Characteristics: Interaction Effects

The main effects of display and task characteristics, while included in the overall model, are
of secondary interest in this study. The main premise of this dissertation is that the effect of visual
display conditions will be moderated by task and individual characteristics. The graph displayed in
Figure 2 indicates that task characteristics will moderate performance in sequential versus

simultaneous display conditions. Degraded conditions are expected to have a more detrimental

40

effect on performance in the sequential display condition. In this section, predictions regarding the
interaction of display type and task characteristics of time pressure, ambiguity, conﬂict, and frame
are discussed.

Display and time pressure. Time pressure is expected to be more detrimental in the sequential
condition, due to greater detrimental effect on effortfui deliberation. If the decision nnakers are
striving to be deliberative and are not given enough time to sequentially analyze all cues and their
interactions, increased error is expected. Simultaneous display of information is expected to induce
a more automatic, less effortfui recognition-based response. The overall impression occurs
immediately, therefore this process is expected to be less vulnerable to effects of time pressure.

One study supporting this expectation compared tine accuracy of recognition-based judgnents
versus explicit deliberation on speeded performance. The group that was explicitly trained was
affected by time pressure, in that performance declined under time pressure. In contrast the group
that was implicitly trained to rely on memory had no difference due to time pressure (Turner &

Fischler, 1993).

Figpre 3. Predicted Effect Of Time Pressure

 

A»

 

 

 

\ﬂ
\' + SEQ
+8110

 

 

ao—u-nou
~<nuqcnn>

 

 

Low TP Mod TP High TP

In a study comparing digital display or a polygon display, Coury and Boulette (1992) trained
research participants to monitor and interpret cues related to the overall state of a hypothetical

system, under varying conditions of time pressure. First, research participants were well trained

41

(384 training trials) and it was noted that decision accuracy was higher for subjects in the digital
display, until the last few training blocks, when performance in botln display conditions were
equally accurate. By the time they were ﬁnished with training, the participants were performing
more quickly in the polygon display condition, with accuracy equal to that of the digital display
condition. Research participants then made system diagnoses under conditions varying in time
stress. Research participants using the digital display were more accurate than research
participants in the polygon display condition when tinerc was no time pressure. However, research
participants in the polygon display condition were more accurate when time pressure was high.

Display and infonnption ambiguity. It is expected that ambiguity will be more detrimental to
the deliberative process, as the missing cues are expected to be more salient in the decisionmaking
process. Deliberate consideration of complex information breaks down when the decision maker
encounters uncertain information (Kahneman, Slovic, & Tversky, 1982; Moser, 1990). Moser
deﬁnes decisionmaking under ambiguity as situations in which the decision maker lacks complete
information regarding relevant facts or outcome probabilities, and points out that decisionmaking
under certainty is quite rare. When the decision object is ambiguous, the subject will have to make
a best guess as to assessment (Gettys, Kelly 111, & Peterson, 1982; Kahneman & Tversky, 1982)
or delay the choice if possible (Frisch & Baron, 1988; Shaﬁr & Tversky, 1992; Tversky & Shaﬁr,
1992).

The proposition that ambiguity will be more detrimental to the deliberative process is based
on an expectation of higher interference within the deliberative process. Ifthe deliberative process
depends on a sequential analysis of information cues and their interactions, the deletion of
particular cues is expected to cause more cognitive distress and higher error than when judgnents

are based on an overall impression. The intuitive process, which considers all cues as a gestalt, is

42

not expected to be as detrimentally affected by the deletion of particular cues, as individual cue

infomnation is not as salient (see Figure 4).

Figpr: 4. Predicted Effect Of Information Ambiguity

 

 

A

 

‘°\—a
\- +sso
+sm

 

 

SOmI-GOU
<03~600>

 

 

Certain Mod AMB High AMB

As ambiguity increase, it is expected to more negatively affects deliberation, such that
under highly ambiguous conditions, decision accuracy will be higher in the simultaneous condition.
Decisionmakers in the simultaneous condition will be more experienced with making judgnents
based on overall impressions, and may not even notice that a few cues are missing. On the other
hand, decisionmakers using a deliberative process will be more aware of missing information, less
able to systematically determine a judgment, and less experienced with nnaking a judgnent based
on an overall impression.

Coury and Boulette (l992)investigated interactions of ambiguity and digital versus polygon
displays, at different levels of time pressure. Research participants using the digital (separable)
display were unaffected by uncertainty until the time pressure was very high. Responses of
research participants using the polygon (integral) display had a more linear relationship with
uncertainty, with increasing amounts of uncertainty leading to more error; however, they performed
more accurately on uncertain targets than research participants in the digital condition when time

pressure was high.

43

Display gpd infprmation conﬂict. Conﬂict among cues is expected to be more detrimental
when cues are presented sequentially. Conscious and deliberate processing of cues will make the
conﬂict more salient. In addition, sequential consideration of cues has already been demonstrated
to result in decision error due to order effects.

In this study, information conﬂict will be presented in a sequential order, in the sequential
display condition. When information is presented sequentially, two types of error have been
reported, where information is weighted more or less depending on the order in which the
information is presented (Anderson, 1981; Hogartin & Einhom, 1992; Jarvenpaa, 1990;
Kahnennan, Slovic, & Tversky, 1982; Slovic & Lichtenstein, 1971). This additional order eﬁ’ect is
expected to result in more error for the sequential display condition (see Figure 5).

Figpre 5. Predicted Effect Of Conﬂict

\Z.‘

 

 

 

 

‘ +sEQ
, +sm

 

 

 

SO—I-OOU
<0.H=ﬂﬂ>

 

 

No Conﬂict Hi Conﬂict

Two order effects have been noted in sequential presentation of information. One is belief-
updating, predicting a general recency effect in that irnforrnation presented last is weighted more
heavily than justiﬁed, particularly in the case of social cognition over time-assessments of
individuals appear to rely most on most recent information (Hogarth & Einhom 1992). This
belief-updating response has been reported for analytical tasks as well (Adelman & Bresnick,

1992). In contrast, a type of primacy effect have also been found (Block & Harper, 1991; Cohen,

44

1993; Payne et al, 1992; Switzer & Sneizek, 1991; Tversky & Kahneman, 1974), where
information presented ﬁrst is more heavily weighted than it should have. A particular example of
primacy is described by Tversky and Kahneman (1974) as anchoring and adjustment. Preliminary
infomnation provides an “anchor” or reference point, from which subsequent information, if
conﬂicting, results in an adjustment of the assessment, but the adjustment is usually not sufﬁcient
for accuracy.

An order effect occurs when the ﬁnal judgnent of decision makers will differ according to the
order that infomnation cues are presented. Hogarth and Einhom (1992) reported several task
characteristics which inﬂuence whether order effects occurring during sequential presentation of
cues is typiﬁed by anchoring and adjustment or belief updating. These autirors found that
anchoring effects were more likely to occur when infomnation cues were complex and when there
was a large number of cues (over 18). It was also more likely to occur when research participants
were requested to make a judgnent after all cues were presented (end-of-sequence), as opposed to
providing a judgnent after each information cue (step-by-step judgnents). Matching these
characteristics to the characteristics of the decision task in this study leads to the expectation of
primacy effects in the sequential display condition.

Display and frame. Kahneman and Tversky (1982) stated that intuitive assessments rely
almost exclusively on singular information; that is, infomnation that is descriptive of the particular
event to be assessed, as opposed to population distribution of events. Consideration of tire
diﬁ’erences between deliberative and intuitive processes leads to the proposition that deliberative
processes are in general, more reactive to the inﬂuence of ﬁaming, that is, the nnanner in which the
decision context is explained.

Research participants in the sequential condition are expected to more effortfnrlly process

information and as a result, are expected to incorporate additional background information, such as

45

past experience, feedback, or background information. Usually, the additional information is
useful, and deliberation is enhanced when considering this contextual information. However, if the
additional information is irrelevant, the deliberative process is disadvantaged. For exarnnplc, when
faced with an ambiguous stimuli such as a written report which is neitlner very good or very bad,
the past performance of the writer may inﬂuence the perception of the quality of the writtern report.
One could say that the "base rate" performance of the writer was considered along with the current

report, in the evaluation of the report.

Figpre 6. Predicted Effect Of Irrelevant Frame Information

 

A

 

 

\ﬁa
+SEQ
+sm

 

 

 

 

:o—e—neu
«ouq:9n>

 

 

No Frame Frame

The assertion that deliberation includes more contextual information is not new. Kahneman
and Tversky (1982) stated that intuitive processes rely on information particular to a decision event
(singular data) to the exclusion of distributional data. Distributional data was deﬁned as
"information that characterizes the outcomes that have been observed in cases of the same general
class." (p 415). This phenomena was listed as a source of error for intuitive decisionmaking, but
there may be instances where the inﬂuence of incidents outside the event of interest may be
detrimental. When the additional infomnation is useful and easily incorporated, deliberative
processes may be improved; however, when the information is independent of the decision event,

deliberative processes, it is expected to introduce more error.

46

Decision context within this proposal is deﬁned as information serving as background or
reference to a speciﬁc decision incident. For example, the outcomes of prior similar decisions
formulates a background for a particular decision. As another example background information
also forms a decision context (T versky & Kahneman, 1981). Ifintuitive processes do not
incorporate contextual information to the same degree as deliberative processes, it is expected that
intuitive processes are more immune to contextual effects.

Kahneman and Tversky (1982) describe the reliance on singular information as a source of
error, but in this study the background information is extraneous to the assessment of information.
It is proposed here that knowledge of previous incidents can serve as a frame which can
subsequently bias the analysis of a decision problem. In this way, knowledge of previous events
can affect deliberative processes more so than intuitive processes. Thus if deliberation is more
likely to incorporate contextual information decisionmakers in the sequential condition are expected

to be more affected by the frame manipulation.

Ingividual Characteristics: Main Effects
ngpitivs ability. Cognitive ability is expected to be positively related to performance in this

task, in both display conditions. It is established that cognitive ability predicts learning and
performance in general, and this task is essentially cognitive in nature. Determination of the
correct response requires perception of cues, and application of mics in order to make correct
decisions. It would be surprising if cognitive ability were not related to performance in tinis task.
Cognitive ability was included as a variable in this study is in order to investigate proposed
interactions between cognitive ability, display condition, and decision context.

Cogpitive _s_tyle. Cognitive style refers to the preference for concrete and speciﬁc facts/mics
versus a preference for concepts, theory, and intuitive processing. Jung (1923) described this

difference as the most fundamental individual difference in distinguishing personality type. Jung

47

described individuals with a preference for concrete facts and mics as “sensing types”, while
individuals with a preference for concepts, theory and intuitive judgnents as “intuitive types.”
Within his infomnation-processing approach to classiﬁcation of personality type, this preference
was regarded as most fundamental in that it describes how individuals perceive their surroundings.
Conﬂict among individual differing in this preference was predicted to be most difficult to
overcome.

Cognitive style is expected to be related to performance as a function of the underlying
cognitive task demand. In this task the demand is for analytical and effortfui processirng, based on
decision mics, which would be more congment with the sensing preferences than the preference for
intuitive, holistic information processing. Thus, congmence of cognitive style and cognitive
demand should be more effective in general. However, an interaction is expected with display
condition and decision context, such that the preference for intuitive processing can be beneﬁcial
when the decision context is degraded.

Individual Characteristics: Interaction Effects

Display and cognitive sbilig. A signiﬁcant interaction is predicted between cognitive ability
and display condition. Individuals high in cognitive ability are expected to perform more
accurately in the sequential display condition. Individuals high in cognitive ability are expected to
more effectively manage the processing demand inherent in analytical decision tasks. Simultaneous
presentation of information is expected to invoke a less effortfui process, based on a holistic
impression of the visual pattern. This impression is expected to be an immediate reaction which
can vary in degree of certainty. The response is not a result of effortfui processing, thus cognitive
ability is not expected to be as inﬂuential in predicting decision accuracy in the simultaneous

condition.

48

Figpre 7. Expected Effect Of Cognitive Ability And Display Condition.

 

A

 

 

 

 

Kﬂ
+SIM

 

 

SO-ﬂ—GOU
(GI-1:00)

 

 

Hi Ability Low Ability

Display and cogpitive ﬁle. A signiﬁcant interaction is expected between decision style and
display condition. Individuals with a preference for analytical processing are expected to perform
more accurately in the sequential display condition than in the simultaneous condition. Individuals
with a high preference for intuitive decisionmaking are expected to perform more accurately in the
simultaneous condition. Individuals with a high preference for intuitive decisionmaking, faced with
a sequential presentation of multiple cues, are expected to be less effective in the deliberative
process, and demonstrate increased decision error associated with deliberation.

F igprg 8. Expected Effect Of Cognitive Style And Display Condition.

 

I;

 

 

 

Kﬂ
+SIM

 

 

 

30—O-ﬂOU
<ﬂI-Icﬂﬂ)

 

 

ﬂ

Hi Ability Low Ability

49

Summary
Figure 9 provides the overall model of predicted relationships discussed witinin this study.
Hypotheses are grouped by (a) main effects of task characteristics (Hi-H4); (b) interactions of task
characteristics and display condition (HS-H8); (c) main effects of individual characteristics (H9-
H10) and (d) interactions of individual characteristics and display condition (I-Il 1-H12).

Figpre 9. Overall model

 

TASK CHARACTERISTICS
Time pressure (H1, 5)
Uncertainty (HZ, 6)

Conﬂict (H3, 7)
Frame (H4, 8)

 

 

 

H5, H6, H7, H8 H1, H2
4 H3, H4

 

 

 

 

 

INFORMATION
DISPLAY DECISION
Simultaneous ACCURACY
Sequential

 

 

 

 

 

H9
H10

 

 

‘7

 

H11, H12
INDIVIDUAL
CHARACTERISTICS
Cognitive ability (H9, 1 1)
Cognitive style (1110, 12)

 

 

 

Task and individual characteristics comprising favorability of decision context are predicted
to have a main effect on decision performance (Hypotheses 1-4; 9-12). The degree of eﬁ’ect will

by moderated by visual display condition (Hypotheses 1,2; 5,6,7,8).

50

Hypotlneses 1 through 4 indicate expected main effects of task characteristics on decision
performance. H1 states that as time pressure increases, performance will decline. Similarly, H2
and H3 predicts that as infomnation ambiguity and information conﬂict is increased, performance
will decline. Similarly, H4 predicts that irrelevant frame information will have a negative eﬁ‘ect on
performance.

The next four hypotheses (1-15 -H8) predict interaction effects between task characteristics and
display condition. The sequential display condition is expected to be more vulnerable to degrading
task characteristics of high time pressure, information ambiguity, information conﬂict, and frame
information. Correspondingly, research participants in the sequential display condition are
expected to perform less accurately than research participants in the simultaneous display
condition when (Hi) time pressure is high, (HZ) information is ambiguous, (1-13) information is
conﬂicting, and/or (H4) irrelevant frame information is provided.

The next group of variables are the individual characteristics expected to have main effects
(H9, H10) on decision performance, that is, cognitive ability and cognitive style. Research
participants who are lower in cognitive ability and/or who have an intuitive decision style are
expected to perform less accurately when conditions are favorable. Research participants with
high cognitive ability and/or analytical cognitive style are expected to perform more accurately in
the sequential display condition when conditions are favorable. Research participants who are
lower in cognitive ability and/or have a preference for intuitive processing are expected to be

perform more accurately in the simultaneous condition.

Chapter 4
METHOD

am le

Research participants consisted of undergraduate students in an introductory course in
management at a large nnidwcstem university. 600 students were recmited, in order to ensure an N
of 500 and thus achieve sufﬁcient (80%) power for statistical analyses (Cohen, 1992; Cohen &
Cohen, 1987; Kraemer & Thiemann, 1987). Incentives to participate in this study consisted of
course credit in lieu of other coursework, and the opportunity to earn money contingent upon task
perfomnance.

Task

The task was a computer-administered team decision task based on simpliﬁed military air
surveillance decisions. Research participants were trained to form judgnents of tlnreat on
hypothetical incoming aircraft based on nine information cues, such as speed, direction, altitude,
etc. This task was created using TIDE2 software (Team Interactive Decision Exercise for Teams
Incorporating Distributed Expertise) for team and individual decision making simulations.
Documentation for the original TIDE2 software can be found in Ilgen and Hollenbcck (1993).

The TIDE2 task enables investigation of decision processes tlnrough multiple presentations of
decision events which require tine individual to make a decision based on up to nine information
cues. Usually, the same cues are considered in each decision event, while cue values are varied.
This allows investigation of processes consistent with the bmnswick lens model of decisionmaking

(answick, 1955). Cue weights can be estimated by regression weights, when there is ﬁnite

51

52

and stable set of cues which vary. In addition, the traditional TIDE2 task is networked such that
several individuals can perfornn as a team, and the software is conﬁgured to capture core
constmcts of the multi-level theory of team decisionmaking.

Research participants were trained to assess the level of threat based on 9 characteristics
of the aircraft. These characteristics include speed, size, angle of ascent/descent, range from base,
altitude, corridor status (whether the aircraft was witlnin a corridor designated safe for civilian
aircraft), direction, radar (hostile to ﬁiendly), and IFF (Identify Friend or Foe) electronic signal.

Research participants completed 180 judgnents, each judgnent taking no more than 20
seconds to complete. Previous pilot tests revealed 20 seconds was considered to be ample time for
decision making by research participants. Most participants completed the session, including the
interactive training, witinin 90 nninutes.

Infonnatipn cues. Research participants played the role of an air patrol ofﬁcer responsible
for monitoring a sector of airspace for incoming aircraft. They were presented with up to eight
visually presented cues and one audio cue representing attributes of each hypothetical aircraft.
There is great contrast between conﬁguration of totally nonthreatening versus the conﬁguration
comprised of all threatening cues. The nonthreatening conﬁguration presents a large green circle
(representing size and friendly Identify Friend or Foe (IFF) signal), far away ﬁ'om the base, within
a safe civilian ﬂight corridor, headed slowly away from the base, at high altitude and ascending,
and emitting ﬁiendiy radar (low pitch slow tempo audio). In contrast, the threatening display
shows a small red circle, very near the base, outside the safe civilian corridor of airspace, heading
straight in at fast speed, at low altitude and descending, and emitting hostile radar (high pitch high
tempo audio) (See Figures 10 and 11). Table 3 describes the nine cues. Each had tinree levels of

tinreat: safe, moderately threatening, and threatening. Figure 12 provides the representations of the

53

Figpre 10. Graphic Display Of Aircraft With All Nontlnreatening Cues.

“beep....beep”

50k

 
 
  
      
  
 

A. “Very on-threatening target”

CUES ASSOCI TED WITH TARGET:

  

Spatial Loc i ~ (Distance to base- the greater the
distance, ~. lower the threat)
Size (Larger is less -
Color of target (Green is Ie .- -
Color of altitude (Green' is less threatening)
Corridor status (inside the dual lines ls safe)
Sound (low pitchlfreq is non threatening)
Arrow (length . speed; short is slow)
(direction I away from base is safe)
Location of dot (if dot is toward top of target
target is ascending, non threatening)

  

54

Fign_rre 11. Graphic Display Of Aircraft With All Threatening Cues.

   
 
 
  
 
    
    

(Base)

8. “Very hreatening target"

CUES ASSOC ‘ TED WITH TARGET:

Size (smaller is m - - threatening)
Color of target (Red is ~ e threatening)
Color of altitude (Red is more -
Corridor status (far outside the dual lines is
most threatening for corridor status)
Sound (high pltchlfreq is threatening)
Arrow (length - speed; long is fast)
(direction - toward base is threatening)
Location of dot (if dot is toward bottom of target
target is descending, threatening)

Table 3

55

Agributes and Rgnges Underlying Perceptusl Cues

 

 

Attribute Description Cue (3 levels)
Speed Miles per hour Length of arrow
Altitude Feet Number, color
Size Meters Size
Angle Degrees: Descent to Ascent. Graphic
IFF Identify Friend or Foe: Megaherz. A radio Color

signal that identiﬁes civilian (low), para-

military, or military (high) aircraft.
Di rection Degrees: Angle of ﬂight ranging from passing Direction of arrow

Co rridor Status

Radar Type

Range

far to the east (+40) to coming straight in (0).

Miles: A corridor is a 20 mile wide lane open
to commercial traﬁic. The status is expressed
in miles from the center of the corridor.

Class: Classes of radar ranging from
weatlner to weapons.

Miles: Distance of the aircraft
Item the operator.

Graphic lines

Auditory

Spatial location

 

Figpre 12. Threat Values Associated Witln Each Cue.

SIZE

LOCATION

ANGLE

SPEED

DIRECTION

lFF

RADAR

ALTITUDE

LOW THREAT

MODERATE THREAT HIGH THREAT

O u...

OUTER RING

O

O MODERATE O SMALL

iMIDDLE RING SECTOR INNER RING

 

Q») Q

 

ASCENDING LEVEL DESCENDING
r a: >
SLOW MODERATE FAST
—_>

[CD

AWAY FROM BASE

 

O M»...

“beep....beep”
LOW PITCH/FREQ

50K

“GREEN”

.\ l.

ANGLED AWAY STRAIGHT IN
“beepabeep” “beep.beep.beep"
MOD PITCHIFREO HIGH PlTCl-IIFREO

5
15k “YELLOW” 5“ k “RED”

57

three levels of tlnreat for each cue. Perceptual cues represented quantitative information. For
example, instead of a number, the size of the aircraft was indicated by a circle which was large,
medium, or small in size depending on the size of the aircraft. Similarly, the range ﬁom the base
was indicated by the location on the another example, tlnree lengths of the arrow representing speed
(short, medium, and long) signify speed (no threat, moderate threat, high threat).

Decision giiematives. Research participants assessed the overall threat of each aircraft, based

 

on seven choices ranging from no threat to very threatening (see Table 4). Each threat level was
associated with a speciﬁc action. For example, if the aircraft was judged as having no threat, the
associated action was "ignore." If the aircraft was judged as having the highest degree of threat,

the corresponding action was "defend", that is, to shoot the aircraft down.

Table 4

The Seven Decision Alternatives

 

(1) IGNORE: The aircraft is assessed as posing no tlnreat. The operator would devote no fnrrther
attention to the aircraft.

(2) REVIEW: The aircraft poses little threat. The operator would leave this aircraft
momentarily, but would return to the aircraft after a short period of time to update its status.

(3) MONITOR: The operator decides to continuously track the aircraft on radar.

(4) WARN: The operator sends a message to the aircraft to steer clear.

(5) READY: The operator sets defensive weapons on automatic. A ship in a readied position is
rarely vulnerable to attack. This stance should not be taken for non-threatening aircraft due to the
possibility of ﬁring mistakenly at innocent aircraft that ﬂy too close.

(6) LOCK-ON: The operator decides to synchronize the ship's radar and attack weapons such
that the weapons are ﬁxed upon the aircraft. An operate may then use oﬁ‘ensive weapons at a

moment's notice. This should be reserved for aircraft almost certain to be threatening.

(7) DEFEND: The operator decides "weapons away" and attacks the aircraft. A defend decision
cannot be aborted once initiated and must only be used when attack by the aircraft is imminent.

 

58

This task does not reﬂect actual military guidelines, but the mles were generated to appear
reasonable. For example, a large aircraft, all other things equal, was described as less threatening
than a small aircraft because ﬁghter aircraft were smaller than civilian airliner aircraft. Perceptual
cues were chosen to be easily interpreted. For example, the cue for aircraft size was a large circle
for large aircraft, and a small circle for small aircraft. As another example, the type of weapons
signal can be civilian, demonstrated by a green color (safe), or military, demonstrated by a red
color (threat).

The mles of engagement required the research participant to consider interactions among the
cues which affect the determination of cue threat. For example, speed and direction make one
interaction, so that one cue could be threatening in itself (high speed) but should not be considered
threatening if the other cue was considered safe (i.e. if direction was safe). Table 5 provides the
mles that guide the correct assessment of each aircraft.

Cue Interpretation Rules

This task has the advantage of a single correct answer for each judgnent, calculated ﬁom a
quantitative representation of the “mics of engagement” (see Table 6). Each cue takes on a
potential threat value ranging from 0 (no threat) to 2 (very threatening). Interactions among the
cues must be considered to ascertain the correct level of threat. The four interactions were
represented by the multiplication of the individual cue values. Thus, if speed is threatening (2) and
direction is nontlnreatening (0), the interaction between spwd and direction would be
nontlnreatening (0 x 2 = 0). The four interactions plus IFF considered alone, determined the threat
of each of the aircraft. The tlnreat level of the aircraft was an additive combination of the threat
level of each cue set. Subjects were trained on the mics, but were not presented with the

mathematical formula.

59

Table 5

 

Rules of Engagement

(1) ASSESSMENT OF IFF (Identify Friend or Foe). All else equal, aircraft with military IFF
(red) are more threatening than ambiguous aircraft (yellow). Civilian (green) aircraft are
nonthreatening for IFF.

(2) ASSESSMENT OF INTERACTIONS. Four rules must be considered in the assessment of
the actual threat of an incoming aircraft:

1. RANGE AND CORRIDOR STATUS - go together, so that CLOSE aircraft that are
WAY OUTSIDE THE CORRIDOR are most dangerous. If either range or corridor status is safe,
the interaction is considered safe. An aircraft that has a threatening range (close) is not a threat if
the corridor status is safe (inside), and vice versa.

2. SIZE AND RADAR - go together, so that SMALL aircraft with WEAPONS radar
(high pitch sound) are most dangerous. If either size or radar is safe, the interaction is considered
safe. An aircraft that has a threatening size (small) is not a threat if the radar is safe, and vice
versa.

3. SPEED AND DIRECTION - go together, so that FAST aircraft coming STRAIGHT
IN are most dangerous. If either speed or direction is safe, the interaction is considered safe. An
aircraft with a threatening direction (straight in) is not a threat if the speed is safe, and vice versa.

4. ALTITUDE AND ANGLE - go together so that LOW-FLYING aircraﬁ that are
DESCENDING are most dangerous. If either altitude or corridor status is safe, the interaction is
considered safe. An aircraft that has a threatening altitude (low) is not a threat if the angle is safe
(ascending), and vice versa.

(3) ASSESSMENT OF THE ENTIRE AIRCRAFT. Assessment of the aircraft as a whole is
based on a quantitative combination of threat assessments of each interaction and the IFF cue.

 

60

Table 6

Underlying Quantitative Model

 

(l) ASSESSMENT OF IFF. (Color of aircraft)

RED = 2 points (military IFF, threat level high)
YELLOW = 1 point (moderate threat)
GREEN = 0 point (civilian, non threatening IFF)

(2) ASSESSMENT OF INTERACTIONS. Four rules must be considered:

A. RANGE AND CORRIDOR STATUS - The threat level of range (0,1,2) is
multiplied with that of corridor status (0, 1,2) to provide a threat level for the interaction (0,1,2,4).
This multiplicative approach applies to all interactions.

B. SIZE AND RADAR - The threat level of size (0,1,2) is multiplied with that of radar
(0,1,2) to provide a threat level for the interaction (0,1,2,4).

C. SPEED AND DIRECTION - The threat level of speed (0,1,2) is multiplied with that
of direction (0,1,2) to provide a threat level for the interaction (0,1,2,4).

D. ALTITUDE AND ANGLE - The threat level of altitude (0,1,2) is multiplied with
that of angle (0,1,2) to provide a threat level for the interaction (0,1,2,4).

(3) ASSESSMENT OF THE ENTIRE AIRCRAFT. Assessment of the aircraft as a whole is
based on the sum of threat levels of each interaction and for IFF. IFF is weighted by 2 so that it
will have a range (0 - 4) equal to that of the interactions, resulting in an overall assessment ranging
from 0 to 20. The correct decision for each aircraft depends on the number associated with the
threat level of the aircraft as a whole, as follows:

IGNORE: 0 - 2MONITOR: 6 - 8 WARN: 9 - 11
REVIEW: 3 - SREADY: 12 - l4 LOCK-ON: 15 -17 DEFEND: l8 - 20

 

Ambiguous decision events. Some of the decision events were ambiguous, that is, research
participants did not see all nine information cues. When the level of threat associated with a
particular interaction was unknown, a rational assessment can only determine a range of "correct"
assessments. For example, if speed was unknown, the threat level for the interaction of speed and
direction cannot be precisely determined. However, a range of “possibly correct” judgments can be

determined.

61

In this study, ambiguous decision events have an underlying correct answer based on all nine
cues, regardless of whether they were available to the decision maker. Research participants were
informed that some decision events will have missing infomation and requested to make their best
guess. They were told that their decision may be consistent with the available information, yet still
be incorrect according to the feedback, because of the unknown information. Scoring was based
on the correct assessment using all nine cues, even if some were not available to the decision
maker, because the focus was on the impact of ambiguity on task performance outcomes as they
can be generalized to realistic settings.

Feedback. After each decision event, participants entered their judgment of overall threat. If
the research participant failed to provide a judgment within the time period a default judgment of
"ignore" was assigned to that aircraft. Once the judgment was entered they were presented with
feedback that provided the correct judgment and the number of points associated with the accuracy
of their judgment.

The feedback screen provided the research participant’s judgment and the actual correct
assessment for the aircraft. Scoring was based on the absolute difference between the research
participant assessment and the correct assessment. Scores range from hit (+2 points), near miss
(+1 point), miss (0 points), incident (-1 point), and disaster (-2 points). If the participant achieved
the correct decision, this was termed a "hit". Thus, if the research participant decided to correctly
"ignore" an aircraft, this was considered a hit. In the same way, if the research participant decided
to correctly "defend“ against an aircraft, this would also be a bit. If the research participant was
one level above or below the correct decision, it was termed a "near miss." Thus, if the research
participant decided to "monitor" when the true decision was "warn", this would be a near miss. If
the research participant was two levels above or below the correct decision, it was termed a "miss".

An example of a miss would be deciding to "ignore" when one should have chosen "monitor."

62

Three levels above or below the correct decision was referred to as an "incident." An example of
an incident would be choosing "defend" when the correct action was "warn." Four or more levels
from the correct decision was termed a "disaster." Thus, if the participant chose to "review" when
the correct decision was "defend" the feedback was that of “disaster” and the research participant

loses 2 points (see Table 7).

Table 7

Scoring Of Accuracy

 

“Hit” Correct threat assessment + 2
“Near Miss” One threat level off + 1
“Miss” Two threat levels off 0
“Incident” Three threat levels off - 1
“Disaster” Four+ threat levels off + 2

 

Research participants were presented with feedback on their decision accuracy immediately
after each decision was made. The fwdback screen also provided the research participant's
performance history (number of hits, near misses, etc.), and a projection of what the person's ﬁnal
total score would be given the performance up to that time. The feedback screen was presented for

3 seconds, followed by the next decision event.

Mrs—Wm

This study has four between-subject variables, two of which were manipulated, and two
which were individual difference variables which were assessed but not manipulated. The two
manipulated between subjects variables were display condition (simultaneous versus sequential

display condition) and decision frame (passive versus aggressive previous error), resulting in four

63

experimental conditions. In addition, there were 3 within-subjects variables (time pressure,
ambiguity, and cue conﬂict) which were consistent throughout all experimental conditions. The
two individual difference variables, cognitive style and cognitive ability, were obtained for each
research participant. The Myers-Briggs Type Indicator was administered to all research

participants and SAT/ACT scores were obtained for each research participant.

Exmrimental Manipulations

Mencing of one presentation. Research participants assigned to the simultaneous

 

condition were presented with all nine cues within a single display, as demonstrated in Figures 4
and 5. Research participants assigned to the sequential condition were ﬁrst presented with a blank
screen followed by cues added one at a time in the same order: range, corridor status, size, radar,
IFF, direction, speed, altitude, angle. When all cues were presented the research participants in the
sequential display condition saw the same screen as the research participants in the simultaneous
condition. Then in both conditions the screen goes blank, allowing research participants an
additional three seconds in which they can make their judgment.

Decision frame. Decision frame was manipulated through background information which
described a recent incident in the same area of responsibility which had disastrous consequences, in
that 200 lives were lost. The manipulation consisted of the type of decision error described for the
incident. In one frame the decision error was one of passivity: a hostile aircraft was assessed as
safe and ignored, with the consequent loss of 200 civilian lives after the hostile attacked the base.
In the other frame the decision error was of over-aggression: a peaceful aircraft was assessed as
threatening and attacked, with the consequent loss of 200 civilian lives.

Each research participant was presented with background information which described the
role they were to play as that of a member of a command-and-control team, recently assigned to be

responsible for an unstable geographic area. The written information described the background

64

scenario. According to the scenario, the area is a part of the middle east with a long history of
political unrest, and the recent incident has made the situation even more unpredictable. The
background information was identical in both conditions except for the description of the disastrous
error made by the previous decision maker.

The error described in both frames resulted in 200 fatalities. The passive ﬁ’ame manipulation
described an error of passivity: a hostile aircraft was not correctly assessed and defended against;
consequently the hostile aircraft attacked and 200 civilians died. In the aggressive frame
manipulation, the decision error was an overly aggressive judgment. In this scenario, the research
participants were told that an aircraft was mistakenly identiﬁed as hostile, and consequently 200
civilians died from the shooting down of an airline aircraft.

In both scenarios, research participants were told that the previous incident should not affect
their current decisionmaking task, which was described as demanding a very objective impersonal
assessment of aircraft information. While this reduces the magnitude of the frame manipulation, it
was important that decisionmakers be focused on achieving objective accuracy.

Time pressure. There were three levels of time pressure: high time pressure (5 seconds),
moderate (10 seconds), and low time pressure (15 seconds). The time allowed for high time
pressure was quite short in order to constrain deliberation. The time pressure was not expected to
constrain the recognition process as Simon (1986) described the recognition process as immediate,
that is, within a few seconds.

After each aircraft was viewed for its amount of time, a blank screen was shown for an
additional 3 seconds. The additional 3 seconds allowed the research participant to view the cues
for the total time, and still be able to send in a judgment. It compensated for the fact that the
sequentially presented aircraft allow the research participants very little time to perceive the last

CDC.

65

_Igf_orr_nation Ambiguity. Ambiguity was manipulated by removal of 2 or 4 information cues.
Certain decision events provided information on all nine cues. Decision events with moderate
ambiguity were missing two cues, each related to separate interactions. Decision events with high
ambiguity were missing four cues, where two of the cues comprise an entire interaction.

Information conﬂict. It was predicted that sequential display of information under time
pressure was more likely to result in primacy error, that is, the infomation presented ﬁrst would be
more inﬂuential in the overall assessment. In order to ascertain whether primacy occurs, decision
events had to be constructed where over-weighting of preliminary information would make a
difference. If subsequent information was consistent with preliminary information primacy could
not be ascertained. Decision events were constructed such that information provided ﬁrst was in
conﬂict with information provided last. Thus, preliminary information may indicate no threat,
followed by information indicating high threat, or vice versa. In this way, if primacy is occurring,
aircraft would be assessed as less threatening than the correct assessment, if the non threatening
information was presented ﬁrst. In the same way, aircraft would be assessed as more threatening
than the correct assessment, if the threatening information was presented ﬁrst.

Construction of decision events. Decision events were generated such that they systematically
vary in time pressure, ambiguity, and cue conﬂict. Table 8 describes the strategy by which
decision events were generated to examine these within-subjects variables. The manipulation and
crossing of (a) three levels of time pressure, (b) three levels of ambiguity, and (c) conﬂict - whether
the ﬁrst three cues were indicative of subsequent cues, resulted in 18 different types of decision
events. A single decision event was constructed for each of the 18 different conﬁgurations of time
pressure, ambiguity, and cue conﬂict. These 18 decision events were presented 10 times in the

same order, a total of 180 decision events assessed by each individual.

66

Table 8

Generation of Decision Events: Within-subject varﬁbles

 

Decision event Time Pressure Ambiguity lst Three
1 l O O
2 2 O O
3 3 O O
4 l l 0
5 2 1 0
6 3 l 0
7 l 2 0
8 2 2 0
9 3 2 O
10 l 0 1
ll 2 O l
12 3 0 l
13 l l l
14 2 l 1
15 3 l l
16 l 2 l
17 2 2 l
18 3 2 l

 

" This chart signiﬁes the manner by which these decision event sets were generated.

The repetition of the 18 decision events was expected to facilitate the recognition process.
Bentin and McCarthy (‘1994) noted that repetition facilitates recognition, and the repetition eﬁ‘ect
occurs primarily when the repetition was immediate. In this study, the 18 decision events were
repeated 10 times in the same order, such that increased recognition of the decision events was

expected, particularly for research participants in the simultaneous display condition.

Procedgg

Research participants were recruited from an undergraduate management class at a
midwestem state university. They were informed that participation in the experiment would

provide extra credit, and would take about three hours of their time. Research participants who

67

expressed a willingness to participate ﬁlled out the Myers-Briggs Type Indicator. Volunteers were
scheduled to three-hour time intervals, in groups of 12-15 at a time.

When research participants reported to the laboratory they w re randomly assigned to
experimental manipulations of display condition and decision frame. They were given written
information describing the task. This background infomnation varied consistent with their assigned
decision frame. Experimental conditions varied randomly over the time period of the data collection
period.

Research participant training. Participants were trained to interpret cue information and
perform the computer task. Training consisted of written information, video training, and
interactive training. The written information included a general overview of the task and
information relating cue values to appropriate judgments. After 5 minutes of studying the written
material, research participants were presented with video training. The video corresponded with
the written script, so that research participants can read along with the video presentation. During
the video presentation, research participants were ﬁrst trained on each individual cue. A visual
representation of the cue or information was presented, followed by the correct assessment. They
were then shown how to operate the computer task. Operation of the computer task was very
simple, as research participants were presented with each aircraft representation and only needed to
enter a judgment with a keystroke.

After the video training, research participants completed a task knowledge questionnaire.
Once the questionnaire was completed, they began the actual task. The ﬁrst ﬁve decision events
were for practice, where research participants were "walked" through the decision task for the ﬁrst
couple of decisions, and the experimenter remains present during this time. Once these practice

decisions were completed the research participant was left to complete the set of decision events.

68

Hypotheses

The following section describes the hypotheses proposed in this study. Hypotheses are
derived from the overall model depicted in Figure 7. The model describes expected effects of task
and individual characteristics on decision performance, as moderated by display condition. Task
characteristics and individual differences were predicted to affect optimality of decision context.
Interactions were expected between these characteristics and display condition.

Task characteristics. The ﬁrst four hypotheses follow ﬁ'om Figure 7 and predict main effects
of task characteristics on decision performance. Time pressure, ambiguity, conﬂict, and ﬂame
were expected to be detrimental to decision performance in general, such that increased time
pressure, ambiguity, and conﬂicting information create a suboptimal decision context.

H1. Time pressure. Time pressure was expected to have a signiﬁcant main eﬂ‘ect on decision
accuracy. As time pressure was increased, decision accuracy was expected to decrease. This
hypothesis was tested using measures AN OVA on mean accuracy scores for each level of time
pressure (O’Brien & Kaiser, 1985; Winer, 1978). The F-test for the main effect of time pressure
was expected to be signiﬁcant at p </= 0.05.

H2. Information ambiguity. Ambiguity was expected to have a signiﬁcant main effect on
decision accuracy. As information becomes less certain due to missing information, decision
accuracy was expected to decrease. This hypothesis was tested using repeated measures ANOVA
on mean accuracy scores for each level of information ambiguity. The F-test for the main effect of
ambiguity was expected to be signiﬁcant at p </= 0.05.

H3. Information conﬂict. Conﬂict of information cues was expected to have a signiﬁcant
main effect of decision accuracy. Conﬂict of information cues was expected to have a negative

effect on decision accuracy. This hypothesis was tested using repeated measures ANOVA, using

69

mean scores for each level of conﬂict. The F -test for the main effect of conﬂict was expected to be
signiﬁcant at p </= 0.05.

H4. Decision frame. A main effect was expected for Frame 1 versus Frame 2 for all research
participants. This hypothesis was tested using ANOVA. The F-test for the main effect of frame
was expected to be signiﬁcant at p </= 0.05.

Interactions between task characteristics and display condition. Interactions were predicted
between display condition and each task characteristic, such that the effect of each task
characteristic was expected to be moderated by display condition. It was expected that research
participants in the sequential display condition would be more detrimentally affected by
unfavorable task characteristics of increased time pressure, ambiguity, conﬂict, and ﬂame.

H5. Time pressure and display. Time pressure was expected to be more detrimental to
research participants in the sequential condition compared to research participants in the
simultaneous condition. Research participants in the simultaneous condition were expected to be
more likely to utilize a recognitional response that does not require as much effortful deliberation,
and thus not be as detrimentally affected by time pressure. This hypothesis was tested using
repeated measures ANOVA. The F-test for the interaction between display and time pressure was
expected to be signiﬁcant at p </= 0.05.

H6. Information ambigpig and display. Ambiguity was expected to be more detrimental to
research participants in the sequential condition, who were more likely to be processing the
infomation sequentially. Research participants in the sequential condition were expected to be
more likely to attempt deliberate consideration of each cue which will make the absence of cue(s)
more salient. Research participants in the simultaneous condition were expected to make a more

holistic judgment, and the absence of a cue(s) was not expected to prevent a holistic impression

70

from occurring. This hypothesis was tested using repeated measures ANOVA. The F-test for the
interaction between display and ambiguity was expected to be signiﬁcant at p </= 0.05.

H7. Lnjpmation conﬂict and display; Conﬂict among information cues was expected to be
more detrimental to research participants in the sequential condition, who were expected to process
the information sequentially and be more susceptible to primacy error. Research participants in the
simultaneous condition were expected to make a more holistic judgment based on all cues seen at
once and the conﬂicting information was expected to be less difficult to integrate. This hypothesis
was tested using repeated measures ANOVA. The F-test for the interaction between display and
conﬂict was expected to be signiﬁcant at p </= 0.05.

H8. Decision ﬂame and display. Research participants in the simultaneous display condition
were not expected to be affected by frame manipulations. However, research participants in the
sequential display condition, were expected to demonstrate a signiﬁcant mean difference in average
decision assessment as a function of frame manipulations. The direction of average judgment was
expected such that research participants in the overly passive condition would provide assessments
that were higher in threat than research participants in the overly aggressive frame condition. This
hypothesis was tested using repeated measures ANOVA. The F-test for the interaction between
display and frame was expected to be signiﬁcant at p </= 0.05.

Individual differences.

Individual differences among decision makers were also predicted to affect decision
performance. Optimal characteristics of the decision maker were predicted to be those who are
high in cognitive ability and have a cognitive style preference that is congruent with the decision
task.

H9. Cogitive ability. Cognitive ability was expected to have a signiﬁcant main effect on

decision accuracy, such that research participants with high cognitive ability were expected to

71

perform more accurately than those with low cognitive ability. This hypothesis was tested using
ANOVA. The F test for the signiﬁcance of cognitive ability as a main effect was expected to be
signiﬁcantly larger than would be expected by chance.

H10. Cogpitive ﬁle. Cognitive style was expected to have a signiﬁcant main effect on
decision accuracy. Individuals with an objective, fact-based decision style were expected to
perform more accurately on this task than those with a preference for intuitive, holistic decision
making. This hypothesis was tested using ANOVA.

Interactions between display and individual characteristics.

Individual differences in cognitive ability and decision style are expected to aﬁ‘ect information
processing, such that low cognitive ability and/or an incongruent decision style should have a
degrading effect on decisionmaking. Consistent with the previous predictions regarding task
characteristics, the simultaneous display condition is expected to be more resistant to degrading
inﬂuences.

H11. _Cpgnitivepbility and di§pI_ay. A signiﬁcant interaction was predicted between cognitive
ability and display condition. Individuals low in cognitive ability were expected to do less well in
the sequential display condition. This hypothesis was tested using ANOVA. The F test for the
interaction between cognitive ability and display condition was expected to be signiﬁcant at p </=
0.05.

H12. 0 itive le and dis la . A signiﬁcant interaction was expected between decision
style and display condition. Individuals with a high preference for “sensing” (fact-based reasoning
as opposed to intuition) were expected to perform more accurately in the sequential display
condition than those with a preference for intuitive processing. This hypothesis was tested using
AN OVA. The F test for the interaction between cognitive style and display condition was expected

to be signiﬁcant at p </= 0.05.

Chapter 5

RESULTS

vervicw

A total of 537 research participants participated in this study, with 286 in the simultaneous

display condition and 251 in the sequential condition. Also, 270 were in frame 1, where the

 

previous error described was overly passive, and 267 were in the frame 2 manipulation, where the

previous error described in background information was one of over-aggression (see Table 10).

Table 9

Between-subjects Variables

 

 

 

 

 

DISPLAY
FRAME Simultaneous Sequential Total
Overly Passive
N Ss. 142 128 270
N Decisions 25,418 22,912 48,330
Overly Aggressive
N Ss. 144 123 267
N Decisions 25,776 22017 47,793
Total
N Ss. 286 251 537
N Decisions 51,194 44,929 96,123

 

 

 

 

 

 

Performance was based on the absolute difference between the true score and the actual

score. The range of this absolute difference spans from 0 to 6. In order to make this accuracy

72

73

score such that a higher number reﬂects higher accuracy, the absolute difference was subtracted
from six, resulting in a number that increases as accuracy increases. The overall mean for this

performance score was 4.69 (s.d. = 1.24), based on a total of 96,123 decisions.

Task Characterigicg
Characterictics of the decision task expected to enhance or degrade the information

processing/decision making process include time pressure (three levels), ambiguity (three levels),
information conﬂict (two lehvels) and decision frame (two categories).

Time pressure: Main effect (H1). Table 10 provides within-subjects ANOVA results for the
impact of time pressure on decision accuracy. All AN OVA analyses reported within this study
were performed using the General Linear Models (GLM) Procedure Repeated Measures Analysis
of Variance procedure of SAS statistical analysis software. For within-subjects manipulations of
task characteristics, such as time pressure, means were calculated for each level, o(i.e. low,
moderate, and high time pressure) for each research participant, and the repeated-measures
AN OVA procedure was run on these means. While there is no difference in the F-value using
either procedure, the sums of squares reported are lower than if the AN OVA was run using scores
for each individual decision.

Table 10

Impact of Tim_e Pressure on Decision Accurm

 

 

Source DF SS MS F Pr Eta2
TIM3 2 2967.04 1483.52 397.90 0.0001 0.43
Error 1072 3996.85 3.72

 

‘ The ANOVA was calculated using means for each participant

74

Table 10, continued.

 

Low TP Mod TP High TP Total
(lSsecsL QOsecsL 45 secsL

 

Mean Accuracy 4.92 4.49 4.68 4.695
SD. (1.08) (1.29) (1.30) (1.24)
N Ss. 537 537 537 1611

N Decisions 32,220 32,220 31,683 96,123

 

 

 

 

 

 

 

6
4.9
4.8
A4]
(34.6
04.5
U
R4.4
A 4.3
C
4.2
Y
4.1

4
Low TP Mod TP High T?

+ TOTAL

 

The effect of time pressure on performance was consistent with predictions. Performance was
expected to be adversely affected by time pressure such that the higher the time pressure the more
detrimental the effect. This is reﬂected in the positive correlation between time pressure and
decision accuracy (r = 0.07; p < .001). However, as reﬂected in the graph, research participants
performed better under high time pressure than under moderate time pressure.

The decision events were examined in case decisions presented under high time pressure were
less difficult than decisions presented under moderate time pressure. Difﬁculty was operationally
deﬁned as deceptiveness: when two interacting cues conﬂict in that one cue is safe and the other is
threatening. The decision rule states that if one cue within an interaction is safe, the entire
interaction should be considered safe regardless of the threat level of the other cue. When one one
is safe and the other is threatening, the aircraft will appear more threatening than it actually is.

This is another variation of information conﬂict; however it differs from the planned manipulation

75

of information conﬂict. The measure of information conﬂict used in this study has preliminary
cues conﬂicting with subsequent cues, thus allowing investigation of order effects in the sequential
presentation condition. Deceptive decision events differ in source of conﬂict: the conﬂict is not
based on order, it is based on cues within a pair of interacting cues. In a deceptive target, there
will be safe and threatening cues, and the target will appear more threatening than it is if judgnent
is based on a sum of safe versus threatening cues. Attention to and application of the decision rule
will result in more accurate and “less threatening” assessments.

As it turned out, there were more deceptive targets in the high time pressure condition. When
deceptive and nondeceptive targets are examined seperately, accuracy was no longer higher under
high time pressure. Findings are further described in the description of exploratory results.

Time Pressure: Interaction with dis la 5 . Research participants in the sequential display
condition were expected to be more negatively affected by time pressure. Research participants in
the sequential condition would then have more accurate assessments for targets presented under
low in time pressure, but would have less accurate assessments than research participants in the
simultaneous condition under high time pressure. This hypothesis was tested using repeated
measures AN OVA. The following table describes ANOVA results for the prediction of accuracy
using display and time pressure as predictors.

Table 11

Impact of Display Condition and Time Pressure on Decision Accuracy

 

 

Source DF SS MS F Pr Eta2
DISP 1 132.77 132.77 11.43 0.0008 0.02
Error 535 6215.59 11.62

TIM3 2 2890.69 1445.34 398.91 0.0001 0.42
TIM3“DISP 2 120.00 60.00 16.56 0.0001 0.02

Error(TIM3) 1070 3876.85 3.62

 

Table 11, cont’d.

76

 

 

 

 

 

 

 

 

 

 

The graph demonstrates the main effect of time pressure was similar for both display
conditions in that research participants performed less accurately in the moderate time pressure
condition. It was predicted that research participants in the simultaneous display condition would
perform better than research participants in the sequential display condition as time pressure

increased; however, the detrimental effect of moderate time was greater for research participants in

 

the simultaneous condition.

Analysis were repeated separately for deceptive and nondeceptive targets. When analyzed

seperately, decision accuracy is no longer higher under high time pressure. Results are described

 

 

 

Time pressure Simultaneous Sequential Total
High (5 secs)
Mean Accuracy 4.67 4.69 4.68
SD. (1.28) (1.32) (1.30)
N 85. 286 251 537
N Decisions 16,874 14,809 31,683
Mod (lOsecs)
Mean Accuracy 4.41 4.58 4.49
SD. (1.32) (1.25) (1.29)
N Ss. 286 251 537
N Decisions 17,160 15,060 32,220
Low (15 secs)
Mean Accuracy 4.90 4.93 4.92
SD. (1.10) (1.06) (1.08)
N 85. 286 251 537
N Decisions 17,160 15,060 32,220
6
4.9
2 4.3
c 2:: .
U 4.5 +SIMULTANEOUS
2 4.4 +SEQUENTIAL
c 4.3
Y 4.2
4.1
4
LOW TP MOD TP HIGH TP

in further detail in the exploratory section. Deceptiveness of target accounts for the lower

accuracy under moderate time pressure; it also accounts for lower performance of subjects in the

simultaneous condition.

Infpmipn ambigpity: Main effect (112). It was predicted that ambiguity from missing

information would degrade the decision context such that research participants would do less well

under ambiguous conditions. The correlation between degree of ambiguity and performance was

positive as expected (r = 0.08; p = .0001), However, while performance was much worse on the

highly ambiguous targets, the performance on the moderately ambiguous targets was better than on

the certain targets.

Table 12

Impact Of Ambigpig On Decision Accuracy

 

 

 

 

 

 

Source DF SS MS F Pr Eta2
AMB 2 2275.78 1137.89 275.64 0.0001 0.34
Error(TIM3) 1072 4425 .34 4.13
Low Moderate High Total
(Certain) Ambiguity Ambiguity
Mean Accuracy 4.72 4.87 4.50 4.69
SD. (1.25) (1.06) (1.36) (1.24)
N Ss. 537 537 537 1611
N Decisions 31,683 32,220 32,220 96,123

 

 

 

 

 

 

Table 12, cont’d

<O)3COO)

Certain

 

Moderate

78

Ambiguous

Decision events were examined for distribution of deceptive targets across levels of ambiguity,

as done previously in the analysis of time pressure. There were more deceptive targets in the

certain condition, and fewer deceptive targets in the moderate condtion. After controlling for

deceptiveness, the effect of ambiguity was as expected. Analyses are further described in the

section describing exploratory analyses.

Information ambiguity: Interaption with display (1:16). It was predicted that research

participants in the sequential display condition would perform less accurately as ambiguity

increased. The next table provides ANOVA results for display, ambiguity, and the interaction

term.

Table 13

Impact of Display Condition and Ambigpig on Decision Accurag

 

 

Source DF SS MS F Pr Eta2
DISP 1 132.77 132.77 11.43 0.0008 0.02
Error 535 6215.59 11.62
AMB 2 2250.27 1125.13 275.97 0.0001 0.36
AMB‘DISP 2 62.87 31.43 7.71 0.0005 0.01
Error(AMB) 1070 3876.85 3.62

 

Table 13, cont’d.

79

 

 

 

 

 

 

 

 

 

 

 

 

Certainty Simultaneous Sequential Total
Low (ambiggous)
Mean Accuracy 4.46 4.54 4.50
SD. (1.36) (1.36) (1.36)
N 83. 286 251 537
N Decisions 17,160 15,060 32,220
Moderate
Mean Accuracy 4.86 4.87 4.87
SD. (1.05) (1.08) (1.06)
N Obs. 286 251 537
N Decisions 17,160 15,060 32,220
High (certain)
Mean Accuracy 4.66 4.79 4.72
SD. (1.30) (1.18) (1.25)
N Ss. 286 251 537
N Decisions 16,874 14,809 31,683
A
c
c
: +SIMULTANEOUS
A +SEQUENTIAL
C
Y

 

CERT MOD AMBIG

Results indicate sigriﬁcant effects of display, ambiguity, and the interaction term on
performance, but not in the manner expected. All research participants performed better on
targets that were moderate in ambiguity. Further, high ambiguity had a more negative inﬂuence on
performance of research participants in the simultaneous condition, when it was predicted it would

have a more negative effect within the sequential condition.

80

Information conﬂict: Main effect 0'13). This hypothesis predicted that research participants

would do less well when cues conﬂicted. However, results indicate no sigiiﬁcant relationship

between conﬂict and performance (r = -.005; p = .30).
Table 14

Impact Of Cue Conﬂict On Decision Accurm

 

 

 

 

 

 

 

 

 

 

 

Source DF SS MS F Pr Eta2
CONF 2 0.98 0.98 0.19 0.67 0.00
Error(CONF) 536 2838.84 5.30
No Conflict Conﬂict
Mean Accuracy 4.70 4.68
SD. 1.36 1.11
N Ss. 537 537
N Decisions 48,330 47,793
5
4.9
A 4.8
g 4.7
u 1::
2 4.4 +rorar
c 4.:
y 4.2
4.1
4
No Conﬂict Conﬂict

It may be that information conﬂict only had an effect on decisionmaking in the sequential

condition, as conﬂict was manipulated as an order effect: information presented ﬁrst conﬂicts with

subsequent information. This is demonstrated in the next analysis.

Another possible explanation is that conﬂict has more effect on the direction of decision error

rather than the amount. In this study half of the conﬂicting decision events presented “safe”

81

information ﬁrst, followed by “threat” information; the other half was in reverse order. The data

were then analyzed for impact of information conﬂict on the response mean. Analyses

demonstrated signiﬁcant relationship between conﬂict and response mean, and are described in the

exploratory section.

Informption conﬂict: Interpction with display (H7) The following table describes the

interaction of display and conﬂict on decision accuracy.

Table 15

Impact Of Display Condition and Infonn_ation Conﬂict On Decision Accura_cy

 

 

 

 

 

 

 

Source DF F Pr Eta2
DISP 1 132.77 11.43 0.0008 0.02
Error 535 6215.59
CONF 1 0.38 0.54 0.00
CONF‘DISP l 40.49 7.74 0.005 0.02
Error(CONF) 535 2798.35
Simultaneous Sequential Total
No Conﬂict
Mean Accuracy 4.65 4.77 4.70
SD. 1.39 1.32 1.36
N Ss. 286 251 537
N Decisions 25,740 22,590 48,330
Conflict
Mean Accuracy 4.67 4.70 4.68
SD. 1.11 1.12 1.11
N 83. 286 251 537
N Decisions 25,454 22,339 47,793

 

 

 

 

 

82

Table 15, cont.

 

+SIMULTANEOUS
+SEQUENTIAL

 

 

 

<O>ZJCOO)

 

NO CONF
CONF

This ﬁnding is consistent with predictions. It was predicted that research participants in the
sequential display condition would perform better than research participants in the simultaneous
display condition when the targets do not conﬂict. It was also predicted that research participants
in the simultaneous condition would do better than research participants in the sequential condition
for conﬂicting targets. While this is not the case, one sees the direction of the ﬁndings consistent
with the predictions, in that conﬂicting targets had a detrimental effect only on decision making in
the sequential display condition.

Decision Frame: Main effect (H4). Frame information was expected impact decision
accuracy and direction of decision error. When the frame manipulation described the previous
error as one of passivity (a hostile aircraft was not defended against when it should have been)
research participants were expected to make somewhat more aggressive assessments. When the
ﬂame manipulation described the previous error as one of over-aggression, research participants
were expected to assess aircraft as less threatening than research participants in the overly passive
frame condition. Results indicate a small but signiﬁcant effect of frame information on decision
accuracy. When the previous error described in the frame was one of agression, accuracy is

higher. This is likely due to an overall tendency of assessing targets as more threatening than they

83

actually are. When the previous error is that of over-aggression, the decision makers were
supposed to be more cautious of agressive assessments. Analyses using accuracy as the dependent
variable are described below, followed by analyses using the response mean as the dependent
variable.

Table 16

 

Impact Of Decision Fme On Decision Accuracy

 

Source DF SS MS F Pr Eta2
FRAME 1 0.27 0.27 4.11 0.04 0.02
Error 535 34.99 0.06

 

 

Passive Agressive

 

Perform 4.68 4.72
SD. 1.25 1.23
N Ss. 286 251

N Decisions 48600 48060

 

 

 

 

 

ZO-M-OMU
<o>mcoo>

 

Passive Aggressive

84

Table 17

Impact Of Decision Frame On Response Mea_n,

 

 

Source DF SS MS F Pr Eta2
FRAME 1 0.04 0.04 0.46 0.49 0.00
Error 535 48.92 0.09

 

Mean Decision Response for Frame 1 and Frame 2

 

Passive Agressive

 

Mean Response 4.02 4.04
SD. 1.77 1.76
N 85. 286 251

N Decisions 48600 48060

 

 

 

 

 

"103201303111”

Z>m§

 

Passive Aggressive

Frame: Interaction with display (ES). The following table describes ANOVA results for the
relative impact of frame and display on decision responses. The graph provides the correct answer

for each target, along with mean responses for research participants in Frame 1 and Frame 2.

85

Table 18

Impact of Display Condition and Frame on Decision Response

 

 

 

 

 

 

 

 

 

 

 

Source DF SS MS F Pr Eta2
DISP 1 234.12 234.12 14.56 0.0002 0.02
FRAME l 6.48 6.48 0.40 0.5258 0.00
D1SP*FR l 0.01 0.01 0.00 0.9843 0.00
Error 533 8571.46 16.08

Passive Agressive
Sequential
Mean Response 3.95 3.97
SD. 1.77 1.76
N Decisions 22,912 22,017
Simultaneous
Mean Response 4.05 4.07
SD. 1.75 1.73
N Decisions 25,418 25776
+SIM
—I— SEQ

<O)ZJCOO>

Passive Aggressive

 

It can be seen that there is minimal difference between the frame conditions. There was no

signiﬁcant interaction between frame and display.

86

Individual Characteristics

Cogpitive ability. SAT and/or ACT scores were obtained where possible on research
participants (N = 375). The following table provides descriptive statistics for the total SAT
(adding verbal and math) and the ACT composite score. Z-scores were also computed.
Table 19

Descriptives for SAT and ACT Scorea

 

 

Variable N Mean SD. Min Max Skew Kurtosis
SAT-Total 139 954.32 145.25 620 1370 0.22 0.14
ACT-Composite 354 22.00 3.30 9 32 -0.29 1.18

 

A total of 114 research participants had scores on both the SAT and ACT measures.
Comparison of these Z-scores indicated that research participants tended to have higher ACT Z-
scores. For example, the Z-scores were computed for all research participants having scores.
When we look at the mean Z-scores for research participants having both scores, the mean ACT 2-
score is 0.19 while the mean SAT Z-score was -0.02. The cumulative percentiles for ACT and
SAT Z-scores were examined, and differences in percentiles were found when equivalent Z scores
of ACT versus SAT were compared. The distribution of the ACT and SAT scores were not the
same (ACT skew = -.29; kurtosis = 1.18; SAT skew = 0.22; kurtosis = 0.14).

ACT and SAT scores were equated using an equipercentile equating procedure. One hundred
and fourteen research participants had scores on both ACT and SAT scores. Using this set of
research participants, data were arranged by cumulative percentile ranks to show the ACT and
SAT scores corresponding to each percentile rank. The computational strategy to generate the
corresponding ACT and SAT scores was based on the identiﬁcation of the observation (ACT, SAT
score) that is closest to that percentile rank. This data was used as a lookup table to identify the

ACT score that corresponds to a particular SAT score at a particular percentile ranking. For

87

example, an SAT score of 620 and an ACT score of 12 were identiﬁed at cumulative percentile of

1; an SAT score of 860 corresponded to an ACT score of 21 (percentile rank of 26); and an SAT

score of 1370 corresponded an ACT score of 32 (percentile rank of 99). This enabled the

transformation of SAT scores to ACT scores for those research participants who did not have

ACT scores. ACT scores were then transformed to Z scores and these scores were used in the

ANOVA analysis.

Table 20

Impact of ACT Z Scoresﬁ and DisplﬂCondition on Decision Accura_cy

 

 

 

 

 

 

 

Source DF SS MS F Pr Eta2
Display 1 135.91 135.91 13.98 0.00 0.03
Z_ACT 1 323.26 323.26 33.26 0.00 0.08
D*Z_ACT l 0.04 0.04 0.00 0.94 0.00
Error 375 3644.55

ACT Score Simultaneous Sequential
Levy.
Mean Accuracy 4.52 4.55
SD. (1.35) (1.33)
N Ss. 26 25
N Decisions 4,680 4,500
Moderate
Mean Accuracy 4.68 4.79
SD. (1.24) (1.20)
N 85. 150 131
N Decisions 27,000 23,580
High
Mean Accuracy 4.77 4.80
SD. (1.19) (1.16)
N 85. 26 20
N Decisions 4,860 3,600

 

 

 

 

 

88

Table 20, cont.

 

+ SIMULTANEOUS
+ SEQUENTIAL

 

 

 

<O>JUCOO>

 

Law G Mod G High G

Results demonstrate a sigiiﬁcant main effect overall on decision accuracy (r = 0.28; p <
0.001) but the interaction between ACT scores and display is not statistically sigiiﬁcant.

Cogpitive sgle. Scores were calculated for the “sensing” and “intuitive” scales of the Myers-
Briggs Type Indicator. These scales are based on forced-choice items and are not completely
independent; however, raw scores were transformed through a normed process as speciﬁed in the
manual. While scores are calculated, the MBTI manual urges the reader not to use them as
quantitative scales; they are for categorization only. The quantitative scores, while suggesting
degree of preference, are ambiguous and cannot be interpreted with precision.

For this study, the scores were transformed as recommended in the manual such that a “0”
indicated no preference; positive scores indicated an intuitive preference, and negative scores
indicated a sensing preference. Research participants were categorized into three groups: (a)
“sensing” types, (b) “intuitive” types, and (e) no decided preference. As can be seen in the table

below, there was no sigiiﬁcant main effect for decision style.

89

Table 21

Descriptives for MBTI Sensing and Intuitive Scam

 

 

 

 

Variable N Mean Std Dev Minimum Maximum

S 493 16.26 7.95 0.00 34.00

N 493 9.79 5.57 0.00 25.00

SN_SCORE 493 1 1.93 26.33 -51.00 67.00

SN_GRP 493 0.99 0.82 0.00 2.00
Table 22

Impact of Display and Cognitive Style on Decision Accuragy

 

Source DF SS MS F Pr Eta2

 

DISP 1 115.26 115.26 9.88 0.002 0.02
SN_GRP 2 15.43 7.71 0.66 0.52 0.00
DISP'SN_GRP 2 6.19 3.09 0.27 0.77 0.00
Error 487 5682.13 11.66

 

5
4.9
4.8
4.7
4.6
4.5
4.4
4.3
4.2
4.1

4

Sensing Neutral Intuitive

~<onqcoo>

 

Cpgpitive style and display. Table 22 also provides ANOVA results for the interaction
between display and decision style. While it appears that individuals with a preference for sensing

performed more accurately in the sequential condition, the interaction was not signiﬁcant.

90

Exploratory Analyses

Several of the hypotheses were not supported by the data. The following section contains
exploratory analyses to investigate post-hoe reasons for results. First, while a main effect for
display was not predicted, results demonstrated a main effect, which is described here. Then, four
issues are explored with further analyses:

1. Manipulation of favorability. The ﬁrst set of analyses in this section addresses the
predicted interactions between display condition and time pressure and uncertainty. The
manipulation of favorability may not have been strong enough. Predicted effects may be more
likely to occur when comparing “highly favorable” versus “highly unfavorable” targets. Therefore,
performance between the two display conditions was compared on targets that were very favorable
(i.e. low time pressure, certain infomation) with targets that were more unfavorable (i.e. high time
pressure, uncertain infomnation).

2. Deceptive targets. Decision events differed in degree of difﬁculty, operationalized as
deceptiveness: Information was classiﬁed as deceptive if within a pair of interacting cues, one cue
is safe and the other cue indicates threat. When one cue is safe, the entire interaction is considered
safe; therefore, deceptive targets may appear more threatening than they actually are. The
distribution of deceptive targets is not equal across conditions of time pressure and ambiguity. It
was found that this unintentional and unequal distribution accounts for the curvilinear results found
for the impact of time pressure and ambiguity on decision accuracy.

3. Perfomnce over time. The third set of analyses investigates the possibility of differences
between the display conditions over time. It was proposed that simultaneous display targets would
elicit a more recognition-based decision process. In this case, the recognition response would be
more likely to occur after the research participants had encountered several sets of targets.

Hypotheses were reanalyzed using performance data on the last 54 out of 185 targets.

91

4. Order effect: Anchoringand-adjustment versus belief-upda_ting. The fourth set of
analyses investigates performance on conﬂicting versus nonconﬂicting targets. Sequential display
of targets was expected to result in less accurate performance when the ﬁrst few cues conﬂicted
with the last few cues. This was supported by the data. Further analyses, presented here,
investigate whether the error was due to anchoring-and-adjustment (ﬁrst cues weighted more) or
belief-updating (more recent cues weighted more) for sequentially presented cue data.

Main effect of display. Display condition varied by sequential versus simultaneous display of
cue information. The main effect of display was predicted to be moderated by time pressure and
ambiguity, therefore no directional hypothesis was made for the main effect of display across all
condtions. While results did indicate that research participants in the sequential display condition
performed more accurately across all conditions, it should be noted that display condition did
interact with task and individual variables such that more accurate predictions of performance can
be attained if display, task, and individual variables are considered.

Table 23

Impact of Display Condition on Decision Accuracy

 

 

 

 

 

Source DF SS MS F Pr Eta2
DISP 1 132.77 132.77 11.43 0.00 0.02
Error(T'IM3) 535 6215.59 11.62
Simultaneous Sequential Total
Mean Accuracy 4.66 4.73 4.69
Std. Deviation 1 .255 1.225 1.242
N observations 51,194 44,929 96,123
N subjects 286 251 537

 

 

 

 

 

 

92

Table 23, cont.

+ Total

 

SEQ SIM

Research participants in the sequential condition had a signiﬁcantly (F = 8.91; p = 0.00 )
higher mean performance score (4.74 > 4.66) across all targets.

1. Manipulation of favorabilig: time pressure and ambigpig For this set of analyses, it was
postulated that expected interaction effects between time pressure and and ambiguity with display
were not as expected because the manipulation of favorability may not have been strong enough.
Time pressure and ambiguity were analyzed separately as representing task-related aspects of
favorability. Conﬂict was not included as conﬂict had no main effect on accuracy. When
variables are analyzed separately, the set of targets in each condition were not fully favorable or
fully unfavorable. That is, targets that were low in time pressure included targets that ranged in
uncertainty and conﬂict. These targets are not completely favorable for deliberation, because of
the targets with high uncertainty. In order to explore fully the effect of optimality, decision events
which were favorable on both task characteristics were compared to decision events which were
unfavorable on all three task characteristics.

Targets which were highly favorable, that is, certain information and low time pressure, were
labeled as “favorable”. Targets which were highly unfavorable, that is, high in time pressure and
ambiguity, were labeled as “unfavorable”. An AN OVA was performed contrasting the favorable

versus unfavorable targets, and any interaction with display.

Table 24

93

Impact of Favorabilityand Display on Decision Accurzgy

 

 

 

 

 

 

 

Source DF SS MS F Pr Eta2
Display 1 0.25 0.25 0.06 0.80 0.00
Favorable 1 4101.64 4101.64 1082.66 0.00" 0.31
Optimal x Display 1 20.40 20.40 5.38 0.02* 0.01
Error (Favorable) 377 9166.56 4.28
Simultaneous Sequential
Highly Favorable
Mean accuracy 5.01 4.96
SD. 0.78 0.70
Moderate Unfavorable
Mean accuracy 4.46 4.37
SD.
Highly Unfavorable
Mean accuracy 4.37 4.27
S .D. 1.26 1.3 1

 

 

 

 

 

 

Fav

Mod Urfav

94

Research participants in the sequential condition performed better than research
participants in the simultaneous condition on the favorable targets, and research participants in the
simultaneous condition performed better than research participants in the sequential condition on
the unfavorable targets, although the effect size was small.

2. Deceptive Targets. Overall, all subjects were most inaccurate for the assessment of low
threat targets. One possible explanation for the lower performance on these low threat targets may
be the deceptiveness of the target. If a target is very threatening all cues are threatening. However,
targets can be made that are very “safe” that contain dangerous cues. This is due to the
interaction rules (if one cue in a paired interaction rule is safe, the entire interaction is safe
regardless of the threat of the other cue), resulting in targets lower in threat more likely to be
deceptive. This can be investigated by detennining the level of deceptiveness in each target. A
target that is not at all deceptive would be one in which either (a) all cues in an interaction are safe,
or (b) there is no safe cue in the interaction. The nondeceptive targets would then be assessed more
accurately if one were using an “equal weighted” strategy, which requires less effort to assess the
threat level.

To investigate the impact of “deceptive” targets, a new variable was created to represent
deceptive (decept = 1) and nondeceptive (decept = 0) targets.

Met of deceptive targets on decision acm The following graph presents mean

decision accuracy for deceptive versus nondeceptive targets, by display condition.

95

Table 25

Impact Of “Deception” By Display Condition

 

 

Source DF SS MS F Pr Eta2
Display 1 6.86 6.86 16.65 0.0001

Deceptive 1 196.15 196.15 785.88 0.0001

Disp x Decept 1 4.06 4.06 16.28 0.0001

Error (Decept) 535 220.39 0.41

 

 

 

NonDec Deceptive

There was a signiﬁcant effect of “deception” on performance, where subjects in both display
conditions were less accurate when the targets were deceptive. Further, subjects in the
simultaneous condition did less well than subjects in the sequential condition when targets were
deceptive. This is not surprising as the deceptive targets required more attention to interactions.
Decisionmakers in the sequential condition had the advantage of having the cues presented in an
order where interacting cues were presented back-to-back, making it easier to attend to
interactions.

The distribution of deceptive targets is not equal across conditions of time pressure and
ambiguity. It was found that this unintentional and unequal distribution accounts for the

cun/ilinear results found for the impact of time pressure and ambiguity on decision accuracy.

96

Deceptiveness. displayaand time pressure. The following graphs illustrate the impact of
deceptiveness as it moderates the effect of time pressure on decision accuracy. Results indicate
that time pressure has no effect on accuracy when targets are not deceptive. When targets are
deceptive, the impact of time pressure is large, and there is no difference between moderate and
high time pressure. The pairing of deceptiveness and time pressure had a larger impact on subjects
in the simultaneous condition.

Table 26

Impact of Time Pressure, Display, and Deceptiveness on Decision Accuracy

 

 

Source DF SS MS F Pr
Display 1 6.86 6.86 16.65 0.0001
Error (Display) 535 220.39 0.41

Tim3 2 211.70 105.85 838.85 0.0001
Tim3 X Display 2 2.66 1.33 10.56 0.0001
Error (Tim3) 1070 135.02 0.13

Deceptive 1 196.15 196.15 785.88 0.0001
Disp x Decept 1 4.06 4.06 16.28 0.0001
Error (Decept) 535 220.39 0.41

Tim3 X Decept 2 331.4 165.70 1125.79 0.0001
Tim3XDecethDisp 2 0.621 0.31 2.11 0.1217
Error (T im3xDecept)1070 157.49 0.15

 

NonDeceptive Targets: Impact of Time Pressure on Decision Accuracy

 

r- 9.0.09.0.“ 999:5???99 9'99”?!"
0 mmumoaanuammwammanuam

W T? MOD HIGH TP

97

Table 26, cont.

Deceptive Targets: Impact of Time Pressure on Decision Accuracy

-O—$M
—D—ﬂm

@9999 999999999 99999
mmwmoadnuammumwmanuem

 

P
O
E
.4
'0

MOD IﬂGHTP

Deceptiveness, display and ambigpity There was also an unequal distribution of deceptive
targets across levels of ambiguity. The following graphs demonstrate the moderating effect of
deceptiveness on the impact of ambiguity and display condition on decision accuracy.
Deceptiveness of the decision event explains the initial ﬁnding that decisionmaking was more
accurate when targets were moderately ambiguous. There were more nondeceptive targets in the
moderately ambiguous condition.

Ambiguity had no effect on decision accuracy when targets were not deceptive, but had a
strong degrading effect on targets that were deceptive. There was no difference between display

conditions when targets were non-deceptive.

Table 27

Impact Of Ambiguity, Deceptiveness, and Display On Decision Accura_cy

98

 

 

Source DF SS MS F Pr
Display 1 4.62 4.62 10.68 0.0011
Error (Display) 535 231.58 0.43
Cert 2 89.01 44.51 204.87 0.0001
CertX Display 2 2.01 1.00 4.62 0.0111
Error (Cert) 1070 232.45 0.22
Deceptive 1 502.26 502.26 1602.82 0.0001
Disp x Decept 1 3.46 3.46 11.04 0.0010
Error (Decept) 535 167.65 0.31
Cert X Decept 2 123.89 61.94 201.83 0.0001
CertXDecethDisp 2 5.81 2.91 9.47 0.0001
Error (CertxDecept)1070 328.40 0.31

 

27A: Non-Deceptive Targets: Impact Of Ambiguity On Decision Accuracy

5.5
5.4
5.3
5.2
5.1

4.9

CERT MOD AMB

HIGH AMB

 

99

Table 27, cont.

27B: Deceptive Targets: Impact Of Ambiguity On Decision Accuracy

+SIM
—D—SEQ

mwmammwwbm

 

CERTAIN MOD AMB HIGH AMB

3. Order effect: Anchorirg-and-adiustment versus belief-updat_ing. Information conﬂict had
a negative impact only on research participants in the sequential display condition.

While greater error is demonstrated for research participants in the sequential condition, it is
not apparent if there is a consistent order bias in terms of primacy or recency. The conﬂicting
targets were examined as to whether decision responses consistently overweighted the ﬁrst few
cues. The next table plots the mean score for conﬂicting targets which had (a) safe cues ﬁrst,
versus conﬂicting cues which had (b) threatening cues ﬁrst.

Table 28

Impact of Order and Display on Decision Accuracy

 

 

Source DF SS MS F Pr Eta2
Display 1 1.09 1.09 1.97 0.16 0.01
Order 1 88.87 88.87 160.45 0.00 0.39
Order x Display 1 4.29 4.29 7.76 0.00 0.03
Error 535 136.58

 

100

Table 28, cont.

Response Means For Targets With First 4 Cues Safe Vs Threatening, By Display

4.2
4.1

4
3.9
3.8
3.7
3.6
3.5
3.4
3.3
3.2
1ST SAFE 1 ST THREAT

 

+ SIM
+850

mmzo'ummau

 

 

 

 

z>mz

It can be seen that research participants in the sequential condition had a lower mean threat
assessment for targets where the ﬁrst three cues were safe, and higher mean threat assessment for
targets where the ﬁrst three cues were threatening. The mean correct response was the same
regardless of order. This supports the expectation of primacy error as a consequence of sequential
presentation of data. The interaction between display and safe vs threat (1st three cues) was
sigriﬁcant for the prediction of response mean.

At the same time results demonstrates what appears to be a primacy effect for decisiomnakers
in the simultaneous condition, though not as strongly as for decisionmakers in the sequential
condtion. Regression analyses were performed to ascertain the relative importance of each cue in
predicting (a) the correct response, (b) decision responses of those in the sequential condition, and
(c) decision responses of those in the simultaneous condition. Results are plotted in the following
table. Also included are the actual weights that were used in the algorithm to assess the correct
answer. Regression weights for the prediction of the correct answer will be somewhat different,
depending on factors such as the degree of intercorrelation among predictor cues. In the algorithm,

the correct answer is determined by equal weighting of the four interactions and the single IFF cue.

 

101

Results indicate that decisionmakers differed from ideal regression weights and from the
algorithm weights. Also, research participants in both conditions had very similar patterns of cue
regression weights, indicating greater weight for cues presented earlier, such as range, corridor
status, the interaction of range and corridor status, size, speed, and direction.

Table 29

R sionWi forDecisionRs ns b Dis la: m risntoId lR in er ts

and Algprim Weights

5 +SIM
45 +880

4 + REGRESS
3.5 +iDEAL

2.5
2

1.5
1

0.5
O

Rge cs Size Rad IFF Spd Dir Alt Ang R9] 82/ Alt] Spl

cs Rd Ang Dir

 

Cue weights are consistent with the prediction of anchoring-and-adjustment for the sequential
condition, along with the previous result that subjects in the sequential condition were more
inﬂuenced in the direction of their mean decision response. However, order eﬁ‘ects cannot explain
the similar ﬁnding for decisionmakers in the simultaneous condition. This may be due to use of a
heuristic based on these characteristics on the assumption they are more important or more
predictive or a visual salience effect (Jarvenpaa, 1990). The ideal regression weights are more

unit-weighted compared to actual weights or to algorithm weights.

Performangc over time

The display manipulation can be critiqued as being a weak manipulation for the elicitation of

recogiition, because the research participants are not so experienced that they immediately

 

102

recogiize each target. One reason for the manipulation as it was performed is that this study
focused on performance error. If research participants were trained to the point where they
recognized targets with case, there would be very little error to compare. Research on implicit
recogntion found effects after one stimulus presentation, and so it was expected that any
advantage due to recogiition of the entire target pattern would appear during the course of the 185
targets that the research participants responded to. In addition, the 185 targets were repetitions (10
sets) of the same 18 targets. Thus, it was expected that any advantage due to holistic recognition
would be demonstrated over the performance of the entire target set. However, it may be that
hypotheses may be more fully supported if we look at performance on the last few repetitions of the
targets.

Table 30

L_ater Perfonnance: Main effect of display

 

 

Source DF SS MS F Pr Eta2
Display 1 8.83 8.83 2.28 0.13 0.01
Error 535 2074.14 3.87

 

Accuracy by Time Period, Display

5
4.9
4.8
4.7
4.6
4.5
4.4
4.3
4.2
4.1

4

EARLY MIDDLE LATE

 

—e— SIM
+550 .

 

 

<O>JUCOO>

 

103

Results indicate the display manipulation had no sigiiﬁcant effect on performance on the last
two sets of targets. Mean performance is provided in the next table, broken out by early (ﬁrst three
sets of targets), middle (sets 4-7), and last three sets of targets. It can be seen that research
participants in the simultaneous display condition were less accurate at ﬁrst, but became as
accurate as research participants in the sequential display condition after several repetitions of the
targets. Next, analyses were run with time pressure and ambiguity included as variables.

Highly favorable versus highly unfavorable targets: Last few repetitions. An ANOVA was
performed using highly favorable (low time pressure, certain information, no conﬂict) and highly
unfavorable (high time pressure, ambiguous information, conﬂict) targets.

Table 31

Impact of Favorability and Display on Decision Accuragz: Last 36 decision events

 

 

 

 

 

 

 

 

 

Source DF SS MS F Pr Eta2
Display 1 0.00 0.00 0.00 0.98 0.02
Favorable 1 1328.75 1328.75 714.99 0.00“ 0.65
Favorable x Display 1 9.86 9.86 5.31 0.02“ 0.01
Error (Favorable) 377 700.62 1.85
Simultaneous Sequential Total
Early 4.47 4.55 4.51
(lst 3 reps) (1.31) (1.28) (1.29)
15444 13554 28998
Middle 4.69 4.77 4.72
(4-7 reps) (1.23) (1.20) (1.22)
20592 18072 38664
Last 4.82 4.88 4.84
(8-10 reps) (1.20) (1.18) (1.19)
15158 13303 28461
Total 4.66 4.73 4.69
(1.25) (1.23) (1.24)
51 194 44929 96123

 

 

 

 

 

 

104

 

 

 

Table 31, cont.
Favorable Unfavorable
Simultaneous
5 .21 3 .76
0.84 1.26
Sequential
5.33 3 .70
0.72 1.26

 

 

 

 

 

 

 

 

A

C

c

U + SIM

R +SEQ

A ._
C

Y

 

FAV UNFAV

As demonstrated above, AN OVA results indicate a sigiiﬁcant interaction between favorability and
display, in support of the overall proposition, and consistent with overall results. Research
participants in the sequential condition performed more accurately when conditions were favorable

than did research participants in the simultaneous condition.

Chapter 6

DISCUSSION

This study drew from three avenues of research (i.e. decisionmakng; automatic versus
effortful cognitive processes; visual cue characteristics) to predict patterns of decision error when 1
complex information is visually displayed. Research in decisionmaking and decision processes

have established systematic pattems of error occur as a consequence of several factors, such as t'

 

limitations of information processing capability (Massaro & Cowan, 1993), expectations and
cogiitive set, susceptibility to framing effects, and reactions to uncertainty or emotional distress
(Parkinson & Manstead, 1992).

Cognitive limitations in information processing capability results in simplifying strategies,
such as the use of heuristics or rules-of-thumb (Stevenson, et al.). A different simplifying
strategy is demonstrated when decisionmakers underutilize relevant prior-probability information
and overweighting similarity when attempting to categorize an array of cues, demonstrating the
error described as representativeness by Kahneman and Tversky (1972; Kahneman, Slovic &
Tversky, 1982). In addition, decisionmakers demonstrate systematic error in processing cues that
are presented over time, by either failing to adjust assessments sufﬁciently as more recent cues are
presented (i.e. anchoring and adjustment error; Kahneman & Tversky, 1982) or by placing too
much weight on the most recent information (i.e. belief updating error; Hogarth & Einhom, 1992).
Research in cogiitive abilities have established primacy and recency effects in basic tasks such as

working memory and retrieval from longterm memory.

105

106

A different type of decision error is demonstrated when ecisionmakers systematically respond
in favor of decision alternatives depending on how the problem is represented. For example,
individuals favor alternatives that are described as a gain rather than a loss, when alternatives do
not diﬁ‘er in any real sense. This would not be a function of limitations in cognitive processing, as
the decision task does not differ in cognitive demand as a function of these alternatives. Instead, it
indicates a systematic bias in favor of alternatives that are positive in phrasing. Biases can also
occur when expectations inﬂuence perception and/or assessment of information. This occurs
regularly as demonstrated by prejudicial attitudes and resistance to information contradictory to
well-established beliefs and opinions. Errors in military tactical decisionmaking have demonstrated I
this, when expectations of peaceﬁrl versus hostile intent inﬂuenced errors of passivity (U .S.S L
Stark), errors of aggression (U.S.S. Vincennes), and errors of ﬁiendly ﬁre (shootdown of ﬁ'iendly
blackhawk helicopters in Iraq).

Decision errors have been assumed to arise from alternative decision processes which are less
effortﬁil and less systematic, compared to an ideal rational decision process based on complete
analysis of all cues and decision rules. Alternative decision processes are often referred to as
intuitive, and assumed when decisionmakers demonstrate increased error. (Kahneman & Tversky,
1982; Kleinmuntz, 1990). At the same time, there are increasing arguments that intuitive decision
processes are more descriptive of actual decisionmaking in complex, dynamic, and naturalistic
settings, and that this intuitive process is adaptive and functional in circumstances where careful
and systematic analysis is not always possible (Zey, 1993).

In this study, the decision task was based on analysis of primarily visually displayed cues
(eight visual cues and one audio cue), requiring assessment of each cue and application of decision
rules to integrate cue information. A review of research on decision processes, decision error, and

visual display characteristics resulted in contradictions for the prediction of decision error.

107

Previous investigations have related intuitive processes to presentation of visual cues as opposed to
quantitative cues. Intuitive decision processes have characteristically been assumed to result in
increased error. Others, such as Hammond (1987) suggested that an intuitive process is
appropriate for decision tasks that have a subjective or aesthetic component. However, in his study
Hammond found that the intuition-inducing condition resulted in high accuracy, even when the task
was analytic in nature.

At the same time, researchers in visual display characteristics appear to be striving to elicit
more intuitive responses, by creating displays that will capitalize on human capabilities for pattern
recogiition. They report high performance in information integration when visual cues are
presented as a pattern, as opposed to more separable cues. However, if visual patterns elicit a
more intuitive response, decision accuracy should be lower than when decisionmakers deliberate
information in a more systematic and effortful manner. Thus, there are competing hypotheses for
the prediction of efficacy when less effortful, more intuitive decision processes are elicited and
utilized by the decisionmaker.

This study sought to clarify apparent contradictions regarding the pattern of deCision error
associated with display of visual cues. First, conceptualizations of intuitive processes were
reviewed to distinguish what is meant by intuition and intuitive error in relation to visual displays.
The notion of intuitive decisionmaking is widely referred to; however, the concept is loosely
deﬁned and alternative conceptualizations differ widely as to the implicit or explicit nature of the
process. The intuitive response elicited by visual displays is best described as resulting from
recognitition of information that has been presented before to decisiomnakers, that is, based on
explicit rather than implicit information. When recogrition is certain, it is automatic, rapid, and
effortless. While certain recognition is explicit and automatic, increased error will arise when the

recogrition response is uncertain and thus more intuitive. It is this error which is of interest in this

 

108

study, and which will be compared to decision error expected to arise from more effortful analysis
of infomnation.

It is not unreasonable to expect an intuitive process to be less accurate when contrasted with
more rational analysis of information. If particular types of visual displays elicit more intuitive
assessments, it would appear that decision accuracy would suffer. However, it has been pointed
out that effortful and systematic analysis is not always possible in dynamic and naturalistic
decision contexts. Intuitive responses are characterized as faster and less effortful, which can be
an advantage when conditions do not allow a careful analysis of information.

Based upon previous ﬁndings (Hammond, 1987; Bennett, 1989) the simultaneous display of a
meaningﬁnl array of visual cues was expected to inﬂuence decision makers to respond more quickly
and intuitvely. This intuitive response was predicted to result in increased error when the decision
task requires integration of complex decision rules, particularly when averaging cue values is not
appropriate. In contrast, sequential and cumulative display of visual cues was expected to
facilitate more effortful consideration of cue values and the decision rules which determine the
correct assessment. In this study, the sequencing of cues further facilitates consideration of
decision rules, as the pairs of cues which interact are presented back-to-back. Thus, the
decisionmaker is guided to systematically consider cues and one interactions.

The expectation that simultaneous presentation of a visual pattern would result in less
accurate integration of information is at least partially counter to reviews in the visual display
literature that report that conﬁgural (pattem-based) displays are better for integration of
information. Separable cues have been reported as less effective for the integration of information,
yet in this study, the sequential display of cues would be perceived as more separable, and is

expected to result in higher accuracy.

 

109

While investigation of main effects of simultaneous versus sequential display of visual cues
may help clarify conditions where conﬁgural versus separable cues are more appropriate, the
primary interest in this study are interactions predicted to occur between visual display
characteristics and variables expected to facilitate or inhibit rule-based deliberation. These
variables determine the favorability of the decision context for effortful analysis of information,
and includes botln task characteristics (e.g. time pressure, uncertainty, conﬂict, and decision frame)
and individual characteristics (cognitive ability and decision style). Sequential display of
infomnation is expected to elicit more accurate decision performance than the simultaneous display
when conditions are favorable, because the decision task demands careful rule-based deliberation
of cues. However, the sequential display condition is expected to result in less accurate decision
performance when the decision context is degraded by factors such as time pressure, ambiguity,
conﬂict, ﬂame, and capability of the decisionmaker. In contrast, simultaneous display of
information was expected to elicit a more recognition-based response which, while not as accurate
as effortful deliberation under favorable conditions, would be more robust when the decision
context is degraded.

Inherent in these predictions are assumptions that the variables chosen in this study as
marnipulations of favorability are in fact degrading to effortful cognitive processing. The task
characteristics chosen for study include time pressure, ambiguity, conﬂicting information, and
irrelevant background (frame) information. In addition, individual difference variables were also
included with the expectation that individual with low cognitive ability and/or an incongruent
cognitive style would also have a detrimental effect on performance. Thus the most favorable
decision making condition would be where there is no time pressure, certain information,

nonconﬂicting infomnation, no irrelevant ﬁame, with decisionmakers who are high in cognitive

110

ability, with an analytical cognitive style. Figure 12 captures the essential expectations of this
study.
Figure 13. Interaction expected between visual display condition and favorability of decision

CODICXI.

 

 

 

+SEQ
+SIM

 

 

 

:Io-a-ooo
~<oan~coo>

 

 

Favorable Degraded

Performance was expected to be less accurate when any of the variables affecting favorability
are degraded.
Main effects.

In general, research participants in the sequential condition performed more accurately
than research participants in the simultaneous condition. This ﬁnding is consistent with
predictions. The congruence between sequential presentation and the underlying cognitive demand
to consider one interactions would account for higher performance in the sequential display.

Results related to decision context variables generally supported expectations, with lower
mean accuracy resulting from several variables associated with favorability of decision context.
Accuracy was signiﬁcantly lower when decisions were made under morderate or high time
pressure. Accuracy was also signiﬁcantly lower when cues were highly ambiguous. Preliminary
analysis indicated no effect of conﬂict on accuracy, but subsequent analysis revealed that conﬂict
signiﬁcantly affected response means. Decision frame also had a signiﬁcant effect on accuracy,

such that the frame eliciting a more passive response (assessing targets as less threatening) resulted

 

111

in higher accuracy. Error in general tended to be characterized by assessing targets as more
threatening than the correct assessment. Cognitive ability had a strong effect on decision
accuracy, as expected. Cognitive style did not signiﬁcantly affect accuracy. Preliminary analyses
for main effects of time pressure, ambiguity, and conﬂict led to misleading results, which were
clariﬁed by subsequent results.

Time pressure, ambiguiig, and deceptiveness. While irnitial results indicated that time
pressure and ambiguity had curvilinear effects on decision accuracy, subsequent analyses
explained these curvilinear effects as due to interaction with target deceptiveness. As described in
the section on exploratory analyses, deceptive targets were deﬁned as targets which appear more
threatening than they actually are, due to presence of several threatening cues combined with a few
safe cues which render cue interactions with the threatening cues as safe. They were generated
randomly as a result of creating decision events representing an equal distribution of the range of
assessment choices.

Deceptiveness of cue information had a strong main negative effect on accuracy, probably
due to increased cognitive demand for consideration of decision rules. Decisionmakers performed
much more accurately when cues within an interaction were consistent, thus allowing an averaging
strategy to be effective. Deceptiveness also accounted for the initial curvilinear nature of the main
effects of time pressure and ambiguity. Time pressure and ambiguity had a strong negative effect
on targets which were deceptive, but not on targets which were not deceptive. The unexpected
ﬁndings regarding deceptiveness demonstrate that visual displays in general can result in rapid yet
accurate response, when decision rules are simple, or when complex rules can be replaced by a
simplifying strategy.

Cue conﬂict. Cue conﬂict was expected to degrade decision making performance. However,

while conﬂict interacted with display condition as predicted, conﬂict had no main effect on

112

performance. This may be due to the nature of targets created with conﬂicting cues. When targets
have both safe and threatening cues, the resulting assessment will be somewhere in the moderate
threat range. Also, because interacting cues were displayed sequentially, there was greater
consistency of interacting cues. For example, the ﬁrst four cues consist of two pairs of interacting
cues. Ifthe ﬁrst four cues are safe, these interactions are easier to assess. These characteristics
of the manipulation of cue conﬂict probably served to reduce the negative effect of conﬂict on
performance in general.

In addition, when conﬂicting targets presented safe cues ﬁrst, the mean threat assessment was 1"

less threatening than the correct assessment and when threatening cues were presented ﬁrst, the

 

mean tlnreat assessment was more threatening than the correct assessment. In this way, the errors r
due to order of presentation cancelled each other out, resulting in a mean that was no less accurate
than nonconﬂicting targets.

ﬂame. Frame had a signiﬁcant main, however, the effect is quite small. Research
participants encountered many (185) decision events. Frame effects may have been washed out
after the ﬁrst few decisions. In addition, decisionmakers were encouraged to provide ebjective
accurate assessments. This was stated within the background infomnation contairning the ﬂame
manipulation, and likely weakened frame effects.

The use of written descriptions to manipulate ﬁ'ame, while consistent with previous
manipulations, is likely too weak to make an impact over a series of decisions. Reading about a
previous error made by another person is not likely to be as salient as feedback regarding one’s
own error. In this case feedback obtained on each decision probably had a much greater impact on
decision performance over time and is likely to be the most salient inﬂuence on subsequent
assessments. However, in realistic and operational settings, a disasterous error is not always made

in the context of a steady stream of decisions. Instead, there may be long periods of inactivity and

113

ambiguity followed by a single ambigous decision event, which may be more vulnerable to frame
eﬁ‘ects from background information and recent events. Further research is needed to investigate
ﬁame effects related to decision errors that occur in realistic settings.

This is not to say that ﬁame effects found in the laboratory setting are not applicable to
Operational settings. Instead, the expectations held by a decisionmaker in realistic and threatening
circumstances is perhaps more likely to affect the judgnent process than any controlled frame
manipulation in the laboratory. For example, in operational military settings, background
information and recent events are additional bits of information relevant to determination of rules
of engagement and tactical decisionmaking. At the same time, operational decisionmakers must be
able to separate assessments that include this background information from assessments of threat
based solely on indicators of threat. Further research is indicated regarding frame eﬁ‘ects as they
relate to single decision events within more operational settings, where background information can
be regarded as relevant to the decision at hand.

In this study, all infomnation cues were perceptual. It is not known whether perceptual cues
are more or less resistant to frame effects. It may be that written descriptions of frame information
has more effect when the decision event is also presented as a written problem. Further research is
indicated to investigate the impact of frame effects on quantitative versus perceptual information
cues. In addition, alternative manipulations of ﬂame should be investigated. When information
cues are perceptual, frame effects may also be more salient when presented perceptually. For
example, in one manipulation the screen may present a number of unknown aircraft at a safe
distance, with one aircraft ﬂying at a more tlnreatening range. The single aircraft would be the
decision event. The alternative manipulation would exclude the other aircraft, or present

alternative frame information, such as a nearby cluster of friendly aircraft.

114

This study demonstrated that frame effects can be associated with interpretation of
visual/audio cues, even when the manipulation is relatively weak. Further research is needed to
understand the dynamics of this effect, to identify situations where the frame eﬂ‘ect would be most
powerful, and to develop interventions to minimize the effect. A useﬁil start would be the
systematic investigation of frame effects (i.e. cognitive set) produced by different manipulations
(e.g. written, verbal, visual, audio, previous error), for different types of decision tasks (e.g.
rational deliberation, recognition, consensus, negotiation) under conditions varying in time
pressure, ambiguity, and consequence of error.

Individuaﬁlifferences. Cognitive ability had a significant main effect on performance, as
predicted. The relationship was somewhat curvilinear, such that the greater diﬁ‘erence was
between research participants low versus moderate in ability, as opposed to moderate versus high
ability. This is probably due to a ceiling effect of the task demand. This nnain eﬁ‘ect is consistent
with the ubiquitous ﬁnding that cognitive ability predicts performance on cognitive tasks.

There was no signiﬁcant effect of cognitive style on decision accuracy. Hypotheses regarding
tlne congruence of cognitive style and task demand/constraints predicted that research participants
with a preference for intuitive tlninking would perform more accurately tlnan research participants
with a preference for logical fact-based thinking when in the simultaneous condition. However,
there was no main effect, nor were interactions with display found. There was no main eﬁ‘ect for
cognitive style, neither was there an interaction with display.

One explanation for the lack of signiﬁcant impact on decision accuracy is that the task
demands between the simultaneous and sequential conditions were identical except for the cue
presentation order. In both cases the cues were visual/perceptual, and correct assessment of threat
was based on an algoritlnm. It may be that the tasks were too similar for any effect of preference to

be demonstrated.

115

In addition, according to Jung (1923), the preference for sensing versus intuition interacts
with three other dimensions, resulting in 16 personality types. These other dimensions may
inﬂuence the impact of cognitive style on decisionmaking. For example, the preference for
extraversion versus introversion describes individual focus of attention, where extraverted
orientation is focused on the outer world, and introverted orientation is more inwardly focused and
reﬂective. The combination of extraversion and sensing preference should enhance the preference
for concrete facts, whereas a combination of introversion and intuitive preference should result in a
more reﬂective and intuitive orientation. Anotlner preference described by Jung is that of
“tlninking” versus “feeling”, where individuals with a thinking orientation prefer to base judgements
on facts and logic, and individuals with a feeling orientation rely more on subjective values,
interpersonal sensitivity, and emotional content. This preference may also inﬂuence cognitive
style, such that the combination of sensing, extraversion, and thinking would be most rational and
fact-based in style, and the combination of intuition, introversion, and feeling would be most
intuitive in style.

Interactions with display condition

Findings regarding main effect expectations were not fully supported; however, it was not the
main purpose of this study to investigate these main effects. The focus of tlnis study was to
investigate whether these degrading variables differentially impact performance depending on the
manner in which information was presented to the research participants. The underlying basis for
these predictions rests on the proposition that information display characteristics can elicit differing
degrees of effortful deliberation versus a more intuitive recognition-based response. Sequential
display of information was expected to elicit more effortful processing of cue information, and thus
be more likely to result in accurate assessments when the decision task demands careful

consideration of complex interacting cues. This advantage for the effortful decision process is

116

expected to break down when tlne decision context is degraded. Thus, while main effects are
predicted in this study, the proposition regarding interactions of display condition with these
degrading variables is the primary focus and contribution of this study. Several of the interactions
with display condition were signiﬁcant and provided partial support for predictions. Signiﬁcant
interaction effects were demonstrated between display and time pressure, ambiguity, and conﬂict.
Display and time pressure. Preliminary analyses indicated that the interaction between
display and time pressure was not as predicted. While subjects in the sequential condition were
expected to perform less accurately under time pressure, they instead performed more accurately 1"

under moderate time pressure. Under high time pressure, there was no difference in accuracy

between display conditions. Subsequent analyses demonstrated that the interaction differed

 

depending on whether the cues were deceptive. When cues were not deceptive decisionmakers
performed equally well regardless of time pressure or display. Consistent with expectations,
subjects in the sequential condition did have higher mean accuracy under low time pressure, and
lower mean accuracy under high time pressure, but the difference was not statistically signiﬁcant.
When cues were deceptive, subjects in both conditions demonstrated much lower accuracy, in
both moderate and high time pressure conditions. Subjects in the sequential condition performed
more accurately than subjects in the simultaneous condition when performing under time pressure.
However, these subjects also performed better than subjects in the simultaneous condition when
targets are deceptive, thus the difference in accuracy between display conditions and time pressure
is due to target deceptiveness. There was however a trend which is consistent with original
expectations, in that decisionmakers in the sequential condition had lower accuracy under high time
pressure compared to moderate, while mean accuracy increased slightly from moderate to high time

pressure in tlne simultaneous condition. It may be that further manipulation of time pressure would

117

demonstrate expected interactions. It is also likely that further time pressure would preclude any
attempt at systematic effortful processing.

Display and ambigpity. Preliminary analyses also indicated that the interaction between
display and time ambiguity was not as predicted. While subjects irn the sequential condition were
expected to perform less accurately under high ambiguity, they instead performed more
accurately. As found with analyses of time pressure, subsequent analyses of the effect of display
and ambiguity demonstrated that the interaction differed depending on whether the cues were
deceptive. When cues were not deceptive decisionmakers performed equally well regardless of L
ambiguity or display, and tended to perform more accurately under moderately ambiguous

conditions. When ambiguity is moderate, therre are fewer cues to attend to, and cues are

 

consistent, thus the moderate ambiguity condition in effect lowered the cognitive processing
required for assessment. When cues were deceptive there was a signiﬁcant interaction that is
partially consistent with expectations. Decisionmakers in the sequential condition performed more
accurately when cue information was certain, but there was no difference in accuracy between
display conditions which were ambiguous. This is partially supportive, in that participants in the
simultaneous condition performed much less accurately on deceptive targets in general, but this
difference is not reﬂected in the ambiguous conditions. Subjects in the sequential condition were
more negatively affected by ambiguity in deceptive targets.

The interactions of time pressure and ambiguity with display condition were not as predicted.
Subsequent analysis demonstrated that the impact of any one variable appears insuﬂ'ncient to
manipulate favorability. Uncertainty may not be suﬁicient to make a decision context unfavorable,
when the other task and individual difference variables are favorable. Data were then reanalyzed

using a combination of time pressure and ambiguity as indicators of favorability.

118

The use of botln variables to irndicate favorability yielded results consistent with predictions.
Research participants in the sequential display cond ition performed more accurately than research
participants in the simultaneous condition when the decision context was very favorable. In
contrast, research participants in the sequential condition performed less accurately than research
participants in the simultaneous display condition for decisions made under very suboptimal
conditions.

While the expected interactions between favorability of time pressure and ambiguity with
display were found to be signiﬁcant, the effect size is quite small. Thus it would appear that no '3'}—

practical signiﬁcance is associated with this interaction. However, another reason for amelioration

 

of expected interaction effects may be the experience level of the research participants. For
example, Coury and Boulette (1992) found signiﬁcant interactions between time pressure,
ambiguity, and digital versus polygon display. Their subjects were more extensively trained (3 84
trials) and only those who reached a 90% accuracy level in the last 100 trials were allowed to
continue. By that time decisionmakers in the polygon condition were performing with equal
accuracy and more quickly. In their study, participants in the digital display condition were more
negatively by time pressure, and by the combination of time pressure and ambiguity, than
participants in the polygon condition. If this study were replicated with more extensive training to
criterion performance in a self-paced context, expected interactions may be more strongly
indicated.

Display and cue conﬂict. The prediction that conﬂicting cues would have a more negative
effect on research participants in the sequential display condition was supported by results. While
research participants in the sequential condition were more accurate than research participants in

the simultaneous condition for both conﬂicting and nonconﬂicting cues, cue conﬂict had a negative

119

effect within the sequential condition. In contrast, cue conﬂict enhanced the perforrnnance of
research participants in the simultaneous condition.

This ﬁndirng is interesting given that the main effect for conﬂict was quite small. Conﬂict was
detrimental, but only to research participants in the sequential display condition. The insigniﬁcant
main eﬂ‘ect was attributed to the fact that conﬂicting targets were by deﬁnitiorn, moderate in threat.
Thus the research participants could eliminate the decision responses of very safe or very
threatening. l

The decision error predicted by order effect biases explains the detrimental eﬂ‘ect of conﬂict
on research participants in the sequential condition. When target cues do not conﬂict, the ﬁrst few

cues presented indicate the general threat of the target. This gives the subject a general impression

 

which is then conﬁrmed by subsequent cue infomnation. However when the targets conﬂict, the
subject encounters disconﬁnning infomnation and must weight the contradictory cues. Research
participants in both the simultaneous and sequential conditions had a tendency to error toward
moderate assessments of threat. However, the sequential display resulted in errors in the direction
of the ﬁrst few cues, as predicted.

Display and frame. The frame manipulation was expected to inﬂuence the direction of
decision error of research participants faced with sequential presentation of information. When the
ﬂame manipulation described the previous error as one of passivity (a hostile aircraft was not
defended against when it should have been), research participants were expected to make somewhat
more aggressive assessments. When tlne frame manipulation described the previous error as one of
over-aggression, subject were expected to assess aircraft as less threatening than research
participants in the overly passive ﬁame condition. This effect was expected to be higher for

research participants in the sequential display condition.

120

The difference in mean judgnent between research participants in the two frame conditions
was not signiﬁcant. While effects of decision frame have been signiﬁcant for single decision
events, it is likely that this manipulation is not powerful enough when research participants are
exposed to many decision events where they get feedback on their own error. The manipulation of
frame through background information is likely to have been washed out by the more proximal
effects of feedback on decisions actually made by each subject.

Display and cognitive ability. It was predicted that subjects with high cognitive ability would
perform more accurately than subjects with low cognitive ability when cues were presented
sequentially, but would not make much difference, if at all, in the simultaneous condition. This
hypothesis was not supported. There was a signiﬁcant main effect of cognitive ability on decision
accuracy regardless of display manipulation. This may be due to the lack of expertise of the
research participants. Research participants were in a [canning mode for the ﬁrst 2/3 of the targets
presented. It is not unreasonable to expect cognitive ability to have a signiﬁcant impact on this
process of acquiring expertise.

It was predicted that a recognition-based response would be more robust with regard to
individual differences in cognitive ability, as effortful deliberation is reduced. In this study, the
manipulation of recognition as a response was not a strong one, in that research participants were
not trained to a threshold level of recognition before data was collected. This was for several
reasons. A primary reason is the focus of this study on pattenns of decision error. If research
participants were experienced to the point that recognition was immediate and certain, very little
error would result, and the impact of display condition would be greatly minimized— research
participants would have the recognition response in both conditions. Instead, the focus was on

ascertairning the impact of these display conditions, and the type of decision errors that can arise as

 

121

a function of these conditions. Thus, training to the point of minimizing error would obviate any
effects from the display conditions.

Display and cognitive style. There was no rrnain effect for cognitive style, neither was there
an interaction with display. As discussed previously, it may be that any effect of preference would
not be demonstrated unless decision tasks are widely different in terms of being based on facts
versus being reliant on intuition. For example, one variable which has been reported to inﬂuence
decision process include quantitative versus perceptual information. In this study, botln display
conditions were exactly alike except for the manipulation of simultaneous versus sequential display
of information. It may be that this manipulation was not strong enough for preferences in cognitive
style to be demonstrated.

Sununapy

Results supported most of the relationships predicted in the overall model. In addition,
subsequent analyses using a more pronounced manipulation of favorability provided further
support, through statistical signiﬁcance and consistency with predicted outcomes. Research
participants in the sequential display condition performed more accurately when performing under
highly favorable conditions (i.e. certain information with low time pressure). They also performed
less accurately than research participants in the simultaneous condition under conditions that were
highly unfavorable. In addition, results demonstrated the ordering effect predicted for research
participants in the sequential display condition

Results indicated relative advantages of the sequential versus simultaneous display of
information, depending on the degree of time pressure and uncertainty inherent in the decision task.
Research participants in the sequential display condition performed more accurately overall, but
were particularly vulnerable to primacy error when the ﬁrst few cues presented conﬂicted with

subsequent cues.

122

The advantages also appear to relate to the congruence of the display condition with the
underlying cognitive demand of the task, which was greater for targets with deceptive cue
interactions. When cue information was consistent such that an averaging strategy can be used,
performance was as good under high time pressure as it was under low time pressure, regardless of
the display condition. When cue interactions were deceptive and required greater deliberation, the
sequential display of information was associated with higher performance.

These results are consistent with the proposition that simultaneous display of information
elicits pattern recognition capabilities which is more robust than effortful cognition under degraded
circumstances. This is not to say that demonstration of intuitive versus effortful cognition was
conclusive; the contribution of tlnis study simply adds to the growing body of research that
describes decision making processes which are alternatives to an objective, rational ideal.

Results were also consistent with a more recent perspective on human decision processes that
is appreciative of the capabilities and advantages of human decision making as opposed to focusing
strictly on the limitations, biases, and errors associated with comparison of actual decision making
to that of a rational ideal. Certainly, it has been demonstrated that human decision making is
associated with consistent tendencies and sources of error. Yet the same tendencies, as sources of
error, may be perceived as advantages if we irnvestigate decision making in more realistic settings,
(i.e. settings which are more ambiguous, complex, and dynamic). While computers and decision
aids can greatly ease the cognitive dennands of adherence to a rational ideal, there are many tasks
for which the human is better suited. In this study, using a complex rule-based decision exercise, a
more intuitive decision process was expected to be (a) less optimal than an effortful striving for
rational processing, but (b) more robust in degraded circumstances, such that the more automatic,

recognition-based process can in fact be an advantage and not simply a source of error. This is

123

consistent with research regarding the use of heuristics and other cognitive strategies that reduce
the cognitive demand of a particular decision scenario (Stevenson & Buscmeyer, 1992).

Implications from this research also relate to applied research in information display. While
it has been stated that presentation of graphic conﬁgurative information (i.e. patterns) can facilitate
integration of infomation, there are linnitations to this proposal, as demonstrated in this study.
First, research participants as a whole performed more accurately, demonstrating superior
integration of cue information, when information was presented sequentially as opposed to the
simultaneous display of a perceptual holistic display. The cue information was more accurately
integrated when cues were presented singly and additionally, as opposed to a simultaneous display
more easily perceived as a whole conﬁgural pattern.

The ﬁnding that one integration was more accurate in tlne sequential display condition ratlner
than the more holistic simultaneous display condition may be due to the underlying cognitive
demand of the decision task used in this study. This task demanded consideration of interactions
among cues such that when one cue is safe, the other interacting cue should also be considered
safe. This can be easily processed with a cue interaction such as speed and direction (fast speed is
not a threat if the aircraft is headed away, as represented by the length and direction of an arrow),
but other interactions were not as familiar, such as range and corridor status. In this situation, the
averaging strategy associated with intuitive decision making can lead to signiﬁcant error in this
decision task. If the targets had been constructed such that an averaging strategy would be
appropriate, the simultaneous condition would not be as handicapping to performance.

Another contribution of this study to applied information display research is the
demonstration of order effects with sequential display of information. While sequential display of
information led to more accurate integration of cue information under favorable conditions, there

was higher error when the conditions degraded. This error was expected to be due to errors of

 

124

primacy, where the ﬁrst few cues presented to research participants would have greater impact on
overall assessment than the subsequent cues. Targets were created to investigate this prediction,
where the ﬁrst few cues were very inconsistent with subsequent cues; for example, when the ﬁrst
three cues are very safe and the last few cues are very threatening. Subject performance was
consistent with this prediction, with research participants in the sequential condition providing
overall assessments which were safer than research participants in the simultaneous condition when
the ﬁrst few cues were safe, and which were more tlnreatening when the ﬁrst few cues were
threatening. Thus, this study revealed the advantages (more precise assessment of complex
information) and disadvantage (a greater tendency for primacy effects when cue information is
inconsistent over time) of sequential display of information.

Subsequent research should investigate this boundary condition of underlying cognitive task
demand. If, as suggested in this study, different decision processes are elicited by a combirnation of
task demand, information display, and decision context, the implication is that decision making
performance can be enhanced through establishing congruence among these factors. For example,
decision perfomnance under soboptimal conditions may be enhanced by change the task demand, by
(a) reducing effortful cognitive load and (b) capitalizing on pattern recognition capabilities. In
addition, research participants can be trained through repetition and training of expertise in order
to achieve a higher degree of recognition-based response. For example, in this study, recognition-
based responses could have been enhanced through repetitive training on safe versus threatening
interactions, followed by repetitive training on overall assessments.

Other researchers also have predicted different decision processes and patterns of error
associated with different display conﬁgurations of information. As the evidence mounts for the
existence and characteristics of these decision processes, information display can become more

sophisticated in enhancing decision performance. For example, researchers have proposed that

 

125

more automatic recognition based responses are desirable, particularly when conditions become
complex or stressful (Mahan, 1992, 1994; Hammond, 1988). At this time, display researchers are
exploring the idea of adaptable display of information, where the manner in which information is
displayed will change, according to factors such as decision context (complexity, ambiguity ,
workload) and individual characteristics (cognitive ability, expertise, fatigue). Before this
human-centered approach to display technology can be realized, we must identify and delineate
more speciﬁcally the characteristics associated with performance under different display
conditions, and identify the explanatory mechanisms for these differences.

Differences in performance between sequential and simultaneous display condition, while
signiﬁcant, were not large. This is probably due to the graphic nature of the cue information in
both display conditions. Larger differences (in performance have been found when display
conditions were manipulations of more ﬁnndamental differences, such as the comparison of display
of numerical data versus the more intuitive, color differentiated graphics used in this study. In this
study, for botln conditions, display was based on perceptual cue information. Sequential cue
information was added to result in a picture that was “built” sequentially. Sequential cues were
presented in fairly fast sequence, particularly under the high time pressure condition. Therefore,
even the research participants in tlne sequential condition were ultimately presented with a
perceptual pattern. The manipulation, in seeking to focus on one aspect of the display type
(simultaneous versus sequential display), was not as extreme a manipulation for the elicitation of a
deliberative versus a recognitional response. The ﬁndings in this study was based on a
conservative manipulation, indicating the potential for greater effect when the display
manipulations are more differentiated.

This study, while not conclusive, indicates the need to study more ﬁnlly the characteristics of

deliberative versus recognition-based decision processes. Issues which call for further

 

126

investigation include the identiﬁcation of display and contextual variables which may elicit one
process over anotlner. In addition, it was assumed in this study that the congruence of the
cognitive demand of the task with deliberation and sequential display would result in higher
performance, which it did. Further research is indicated to verify the extent to which the task
demand should be congruent with display characteristics in order to establish principles for

maximization of decision performance.

 

Chapter 7

REFERENCES

Adelrnan, L., & Bresnick, T. (1992). Examining the effect of irnfomnation sequence on Patriot Air
Defense ofﬁcers' judgments. Organizational Behavior and Human Decision Proce§§e_s, 204-
229.

Agor, W. H. (1989). What is intuition? In W. H. Agor (Ed), Intuition in organizations: Leading
and managing productively. Newbury Park: Sage Publications.

Allwood, C., & Montgomery, H. (1987). Response selection strategies and realism of conﬁdence
judgnents. Organizational Behavior and Human Decision Processes, 3, 365-383.

Anderson, N. H. (1981). Foundations of InformLion Integratio_n. New York: Academic Press.

Andre, A. D. & Wickens, C. D. (1992). Compatibility and consistency in display-control systems:

Implications for aircraft decision aid design. Human Factors, 34 (6), 639-653.

 

Ashby, F. G., & Maddox, W. T. (1990). Integrating infornnnation from separable psychological
dimensions. Journal of Experimeml Psychology: Hmn perception and Perfomnm, ﬂ
(3), 598-612.

Bargh, J. A. (1989). Conditional automaticity: Varieties of automatic inﬂuence in social
perception and cognition. Unintended thought (pp 3-51). New York: Guilford Press.

Barnard, C. (193 8). The Functions of the Executive. Cambridge, MA: Harvard University Press.

Bar-Hillel, M. (1982). Studies of representativeness. In D. Kahneman, P. Slovic, and A. Tversky
(Eds) Luggment under uncertaiaty: Heuristics_ and bia_sas. Cambridge: Cambridge

University Press.

127

 

128

Barnett, B. J ., & Wickens, C. D. (1988). Display proximity in multicue infomation integration:
The beneﬁt of boxes. Human Factors, ﬂ, 15-24.

Beach, L. R. (1993). Broadening the deﬁnition of decision making: The role of prechoice screening
of option. Psychological Science, 4, 215—220.

Beach, L. R. (1990). l_n_gnge theory: Decision making in personal and organizational contem. New
York: John Wiley and Sons.

Beach, L. R., & Lipshitz, R. (1993). Why classical decision theory is an inappropriate standard for
evaluating and aiding most hnunnan decision making. In G. Klein, J. Orasunu, R
Calderwood, & C Zsambok (Eds.) Decision making in action: Models and methgs.
Norwood, NJ: Ablex Publishing Corporation.

Beach, L. R., & Mitchell, T. R. (1978). A contingency model for the selection of decision
strategies. Academy of Management Review, 3, 439-449.

Bennett, K. B., & Flach, J. M. (1992). Graphical displays: Implications for divided attention,
focused attention, and problem solving. Human Factors, 3 (5), 513-533.

Bentirn, S., & McCarthy, G. (1994). The effects of immediate stimulus repetition on reaction time

and event-related potentials in tasks of different complexity. qu mal of Exparimental

Pschology: Leanning, memog, and cogpition, 20(1), 130-149.

 

Block, R A., & Harper, D. R. (1991). Overconﬁdence in estimation: Testing the Anchoring-and-
Adjustment hypothesis. Organizational Behavipr and Human Decision Processes, Q, 188-
207.

Boiney, L. G. (1993). The effects of skewed probability on decision making under ambiguity.
Organizational Behavior and Human Decision Processes, 16, 134-148.

Boles, D. B. & Wickens, C. D. (1987). Display formatting in infomnation integration and

nonintegration tasks. Human Factors, _2_9(4), 395-406.

129

Boulette, M. D., Coury, B. G., & Bezar, N. A. (1987). Classiﬁcation of multidimensional data
under time constraints: Evaluating digital and conﬁgural display representations. 1_n

Proceedings of the Human Factors Society 32nd Meeting (pp. 116-120). Santa Monica, CA:

 

Human Factors Society.

Brunswick, E. (1955). Representative design and probabilistic theory in a functional psychology.
@choloniew. 62. 193-217.

Carswell, C. M. & Wickens, C. D. (1987). Information integration and the object display: An :-
interaction of task demands and display superiority. Ergonomics 30 511-527.

’—9

Carswell, C. M., & Wickens, C. D. (1990). The perceptual interaction of graphical attributes:

Conﬁgurality, stimulus homogeneity, and object integration. Perception Qa Pachophysics, %
3‘

ﬂ, 157-168.

 

Casey, E. J. (1986). Visual display representation of multidimensional systems: The eﬁ'ect of
information correlation and display integrality. In Proceedipggnf the Human Factors Society
30th Annual Meeting (pp. 430-434). Santa Monica, CA: Human Factors Society.

Cohen, J. (1992). A power primer. Psychological Bulletin. 112, 155-159.

Cohen, J ., & Cohen, P. (1987). Applied multiple mession/correlation aﬂlysis for the beh_a_vi_o_ra_l

sciences. Hillsdale, NJ: Lawrence Erlbaum Ass.

 

Cohen, M. S. (1993). Three paradigns for viewing decision biases. In G. Kleirn, J. Orasunu, R.
Calderwood, & C Zsambok (Eds.) Decision making in action: Models and methods.
Norwood, NJ: Ablex Publishing Corporation.

Cohen, M. S. (1993). The naturalistic basis of decision biases. In G. Klein, J. Orasunu, R.
Calderwood, & C Zsambok (Eds.) Decision making in action: Model‘s_ and methods.

Norwood, NJ: Ablex Publishirng Corporation.

 

130

Cooper, L., Schacter, D., Ballesteros, S., & Moore, C. (1992). Priming and recognition of

transformed tlnree-dirnensional objects: Effects of size and reﬂection. Journal of

Exmrimental Psycholqu: Learning, Memog, and Cogpition, ﬂ, 1, 43-57.

 

Coury, B. G., & Boulette, M. D. (1992). Time stress and the processing of visual displays. Human

Factors, 3(6), 707-725.

 

Coury, B. G., Boulette, M. D., & Smith, R. A. (1989). Effect of uncertainty and diagnosticity on
classiﬁcation of multidimensional data with integral and separable displays of system status.
Human Factors, 3_1 (5), 551-569.

Coury, B. G., & Pietras, C. M. (1989). Alphanumeric and graphic displays for dynamic process
monitoring and control. Ergonomics, 32, 1373-1389.

Curley, S. R, Yates, F ., & Abrams, R. A. (1986). Psychological sources of ambiguity avoidance.
Organizational Behavior and Human Decision Prﬂsaes, 38, 230-256.

Dawes, R. M. (1982). The robust beauty of improper linear models in decision making. In D.

Kahneman, P. Slovic, and A. Tversky (Eds.) Judgment under uncertainty: Heuristics and

biases. Cambridge: Cambridge University Press.

 

Denes-Raj, V., & Epstein, S. (1994). Conﬂict between intuitve and rational processing: When
people behave against their better judgnent. Journal of Personality and Soaial PsychoLogy,
_6__§, 819-829.

Edwards, W. (1954). The theory of decision making. Psychological Bulletin, §_1_, 38-417.

Einhom, H. J., & Hogarth, R. M. (1985). Ambiguity and uncertainty in probabilities inference.
Psychological Review, 22 (4), 433-461.

Einhorrn, H. J ., & Hogarth, R. M. (1981). Behavioral decision tlneory: Processes of judgnent and

choice. Annual Review of Psychology, 32 53-88.

_9

131

Epstein, S. (1994). Integration of the cognitive and the psychodynamic unconscious. American
Psychologist, 42(8), 709-724.

F iske, S. & Taylor, S. (1991). Social cognition. In S. T. Fiske and S. E. Taylor (Eds.) Soc_ial_
Qgpitio_n. New York: Mcgraw-Hill Inc.

Frisch, D. (1993). Reasons for ﬂaming effects. Organizational Behavior and HumaLDecision
Processes. 54. 399-429.

Frisch, D. & Baron, J . (1988). Ambiguity and rationality. Journal of Be@ oral Decision Making,
1, 149-157.

Garner, W. R. (1978). Selective attention to attributes and stimuli. Journal of Experimental
Psychology: General, 107. 287-308.

Gettys, C., Kelly 111, C., & Peterson, C. (1982). The best-guess hypotlnesis in multistage inference.
In D. Kahnemarn, P. Slovic, & A. Tversky (Eds.) Judgment under uncertainty: Heuristics
and biases. Cambridge, UK: Cambridge University Press.

Gilliland, S. W., Schmitt, N., & Wood, L. (1993). Cost-beneﬁt determinants of decision process
and accuracy. Organizational Behavior and Human Decision Processea, §§, 308-330.
Goldberg, P. (1989). The many faces of intuition. In W. H. Agor (Ed.) Intuition in Organizatigns:

Leading and managing productively. Newbury Park: Sage Publications.

Hammond, K. (1955). Probabilistic functionalism and the clinical method. Psychological Review,
62, 255-262.

Hammond, K. (1993). Naturalistic decision making from a Brunswickian viewpoint: Its past,
present, future. In G. Klein, J. Orasunu, R. Calderwood, & C Zsambok (Eds.) D_ec_i§jpa

making in action: Models and methods. Norwood, NJ: Ablex Publishing Corporation.

 

132

Hammond, K., Hamm, R., Grassia, J ., & Pearson, T. (1987). Direct comparison of the efﬁcacy of
intuitive and analytical cognition in expert judgment. IEEE Transactions on Systems. Ma_n_,
and Cybernetics. l_7(5), 753-770.

Hammond, K., Harvey, L., & Hastie, R. (1992). Making better use of scientiﬁc knowledge:
Separating truth from justice. Psychological Science, 1, (2), 80-87.

Hermstein, R. J. (1990). Rational choice theory: Necessary but not sufﬁcient. American
Psychologist, 4_5_, 356-367.

Hogarth, R. M. (1975). Decision time as a function of task complexity. In Wendt & Vlek (Eds.),
Utilig, probability. and human decision mgipg (pp. 321-338).

Hogarth, R. M. (1981). Beyond discrete biases: Functional and dysfunctional aspects of
judgmental heuristics. Psychological Bulletin, 20, 197-217.

Hogarth, R. M., & Einhom, H. J. (1992). Order effects in belief updating: The belief-adjustment
model. Cognitive Psi/chow. 21. 1-55.

Ilgen, D. R., & Hollenbcck, J. R. (1993). Effective team performance under stresaand nonna_l

conditions: An experimental paradigm, theou and data fro studyaa' g team decisipn making in

 

hierarchical teams with distributed expertise. Technical Report No. 93-2. Arlington, VA:
Ofﬁce of Naval Research.

Jacob, V., Gaultrney, L., & Salvendy, G. (1986). Strategies and biases in human decision making
and their implications for expert systems. Enaviour and Information Techmgy, 5(2), 119-
140.

Jarvenpaa, S. L. (1990). Graphic displays in decision making: The visual salience effect. Joarnal
of Behavioral Decision Mam, 2, 247-262.

Jung, C. (1923). Psychological Types. New York: Harcourt Brace.

 

133

Kahn, B. E. & Sarin, R. K. (1988). Modeling ambiguity in decisions under uncertainty. Journal of
Consumer Reseamh, Q, 265-272.

Kahnemarn, D. (1991). Judgnent and decision making: A personal view. Psychological Science, 2,
142-145.

Kahneman, D., Slovic, P., & Tversky, A. (1982). Judgment under uncertainty: Heuristicaang
b_iaae_§. Cambridge, UK: Cambridge University Press.

Kahneman, D., & Tversky, A. (1982). Intuitive prediction: Biases and corrective procedures. In D.
Kahneman, P. Slovic, & A. Tversky (Eds.) Judgment under uncertainty: Heuristics and
bia_s_e_s. Cambridge, UK: Cambridge University Press.

Kahneman, D., & T mversky, A. (1982). On the study of statistical intuition. Cogpition, 11, 123-

 

141.

Kahneman, D., & Tversky, A. (1972). The psychology of preferences. We, 161-173.

Kerstholt, J. (1992). Information search and choice accuracy as a function of task complexity and
task structure. Acta Pscyhologica, &, 185-197.

Kleirn, G. (1993). A recognition-primed decision (RPD) model of rapid decision making. In G.
Klein, J. Orasunu, R. Calderwood, & C Zsambok (Eds.) Decision making in action: Models
and methods. Norwood, NJ: Ablex Publishing Corporation.

Klein, 0., Calderwood, R., & Clinton-Cirocco, A. (1988). R_apid decision making on the
ﬁreground (ARI Technical Report No. 796). Alexandria, VA: US. Army Research Institute
for the Behavioral and Social Sciences.

Klein, G., Orasunu, J ., Calderwood, R., & Zsambok, C. (1993). Decision making in action:
Models and methods. Norwood, NJ: Ablex.

Kleinmuntz, D. (1990). Why we still use our heads instead of formulas: Toward an integrated

approach. Psychological Bulletin. 107. 296-310.

 

 

134

Kleinmuntz, D. (1985). Cognitive heuristics and feedback in a dynamic decision environment.
Management Science, 3_l(6), 680-702.

Kosslyn, S. M., & Anderson, R. A. (1992). Frontiers in cogpitive neuroscience. Cambridge: MIT
Press.

Kraemer, H., & Thiemann, S. (1987). How many research paﬂicipants? Newbury Park: Sage
Publications.

Logan, G. D. (1988). Automaticity, resources, and memory: Theoretical controversies and
practical implications. Human Factors, 39 (5), 583-598.

Lord, R. G. & Maher, K. J. (1991). Cognitive tlneory in industrial and organizational psychology.

 

In M. Dunnette and L. Hough (Eds.) Handbook of Industrial and Organizational Psychology,

Vol. 2. Palo Alto, CA: Consulting Psychologists Press.

Maharn, R. P. (1994). State-depenedent cognitive ﬁnnctioning: Implications for work behavior.
Human Performaaaa, 1(2), 81-83.

Mahan, R. P. (1994). Stress-induced strategy shifts toward intuitive cognition: A cogpitive
_caatinuum framework approach. Human Performance, 1(2), 85-118.

Mahan, R. P. (1992). Effects of task uncertainty and continuous performance on knowledge
execution in complex decision making. International. Journal of Computer Integrated
Manufacturing, 3(2), 58-67.

March, J. G., & Shapira, Z. (1992). Behavioral decision theory and organnizational decision theory.
In M. Zey (Ed) Decision making: Alternative to rational choice models. Newbury Park:
Sage Publications.

Massaro, D. W., & Cowan, N. (1993). Information processing models: Microscopes of the mind.

Annual Review of Psychology, 44 383-425.

—9

135

Matin, E., & Boff, K. (1988). Information transfer rate with serial and simultaneous visual display
formats. Human Factors. 3_0(2), 171-180.

Meehl, P. E. (1957). When shall we use our heads instead of the formula? Journal of Counseling
Psychology, 4, 268-273.

Mintzberg, S., Raisinghani, D., & Theoret, A. (1976). The structure of "unstructur " decision
processes. Administrative Science Ou_art_e;ly, 2_l_, 246-275.

Mitchell, T. R. & Beach, L. R. (1990). "... Do I love thee? Let me count..." Toward an
understanding of intuitive and automatic decision making. Omanizational Behavior and
Human Decision Processes. 41, 1-20.

Moser, P. K. (1990). Rationality in action: general introduction. In P. K. Moser (Ed) Rationality
in action: Contemporapy approaches. Cambridge: Cambridge University Press.

Musen, G. (1991). Effects of verbal labeling and exposure duration on implicit memory for visual
patterns. Journal of Experimen_tal Psychology: Learning, memog, Ed cogpition l_7(5), 954-
962.

Nisbett, R. E., & Ross, L. (1980). Humgm inference: Strategies and shortcomings of social
judgment. Englewood Cliffs, NJ: Prentice-Hall.

O’Brien, R. G. & Kaiser, M. K. (1985). MANOVA Metlnod for analyzing repeated measures
designs: An extensive primer. Psychological Bulletin, 21(2), 316-333.

Onken, J., Hastie, R., & Revelle, W. (1985). Individual differences in the use of simpliﬁcation
strategies in a complex decision-making task. qumal pf Experimental Paychology: Hum
Perception and Perfomnm, l_1_(1), 14-27.

Orasunu, J ., & Connolly. (1993). The reinvention of decision making. In G. Klein, J. Orasunu, R.
Calderwood, & C Zsambok (Eds.) Decision making in action: Models and methods.

Norwood, NJ: Ablex Publishing Corporation.

 

 

136

Parkinson, B. & Manstead, A. S. (1992). Appraisal as a cause of emotion. In M. S. Clark (Ed)
Ma. (13 in the Review of Personality and Social Psychology series). Newbury Park:
Sage.

Payne J. W., Bettrnan, J. R. & Johnson, E. J. (1992). Behavioral decision research: A constructive
processing perspective. Annaal Review of Psychology, 4_3_, 87-131.

Payne, J. W., Bettrnan, J. R, & Johnson, E. J. (1988). Adaptive strategy selection in decision
making. Journal of Experimental—Psychology: Learning, Memog, and Cogpition, 1_4, 534-
552.

Pomerantz, J. R, & Pristach, E. A. (1989). Emergent features, attention, and perceptual glue in
visual form perception. Jourmnl of Experimental Psychology: Human Perception and
Performance, 3, 422-435.

Reber, A. S. (1993). Implicit Learning and Tacit Knowledge. Oxford Press.

Rowan, R (1989). What it is. In W. H. Agor (Ed) Intuition in organizations: Leading and
managing productively. Newbury Park: Sage Publications.

Sanderson, P. M., Flach, J., Buttgieg, M., & Casey, E. Object displays do not always support
better integrated task performance. Human Factors. 1989, 3_1 (2), 183-198.

Savage, L. (1954). The Foundations of Statisﬁa. New York: Wiley.

Schacter, D. L. (1987). Implicit memory: History and current status. Journal of Experimental
Psychology: Learning- Memory. and Cognition. 13, 501-518.

Schneider, W. & Shiffrin, R. (1977). Controlled and automatic human infomnation processing: 1.
Detection, search and attention. Psychological Review, 8_4 (1), 1-66.

Seger, C. A. (1994). Implicit Leanning. Psychological Bulletin, _l_l§(2), 163-196.

Shaﬁr, E., & Tversky, A. (1992). Thinking tlnrough uncertainty: nonconsequential reasoning and

choice. ngnitive Psychology. 24, 449-474.

137

Shiﬂ‘rin, R. & Schneider, W. (1977). Controlled and automatic human information processing: II.
Perceptual learning, automatic attending, and a general theory. Psychological Review, 34(2),
127-190.

Simon, H. A. (1992). What is an "explanation" of behavior? Pavehologiaaﬂcience. 3(3), 150-
161.

Simon, H. A. (1990). Invariants of human behavior. AnnuaL Review of Psychology, 4_1_, 1-19.

Simon, H. A. (1989). Making management decisions: The role of intuition and emotion. In W. H.
Agor (Ed) Intuition in organizations: Leading and managing productively. Newbury Park:
Sage Publications.

Simon, H. A. (1986). The infomnation processing explanation of gestalt phenomena. Computers in
Human Behavior. 2, 241-255.

Simon, H. A. (1955). A behavioral model of rational choice. anrterly Joml of Economica, _62,
99-118.

Slovic, P., Fischhoff, B., & Lichtenstein, S. (1977). Behavioral decision tlneory. Annual Revim of
Psychology, 23, 1-39.

Slovic, P., & Lichtenstein, S. (1971). Comparison of bayesian and regression approaches to the
study of information processing in judgnent. Organizational Behavior and Hm
Performance, 9, 649-744.

Stevenson, M. K., Buscmeyer, J. R, & Naylor, J. C. (1990). Judgnent and decision nnaking
theory. In M. D. Dunnette and L. Hough (Eds.) Handbook of Industrial and Organizational
Psychology, Volume 1. Palo Alto, CA: Consulting Psychologists Press, Inc.

Switzer, F. & Sniezek, J. (1991). Judgnent processes in motivation: Anchoring and adjustment
effects on judgnent and behavior. Organizational Behavior and Human Decision Processes.

42, 208-229.

138

Thibaut, J ., & Walker, L. (1978). A theory of procedure. California Law Review, Q, 541-566.

Treismarn, A. M. (1986). Properties, parts, and objects. In K. Boff, L. Kauﬁnnann, and J. Thomas
(Eds.), Handbook of Perception and Human Peder-mange (pp. 35-1-35-70). New York:
Wiley.

Turner, C. & Fischler, I. (1993). Speeded tests of implicit knowledge. Journal of experimengal
psvcholqg: Learning. memory. and cognition. 12(5), 1165-1177.

Tversky, A. (1972). Elinnination by aspects: A theory of choice. Psychological Review, 2(4), 281-
299.

Tversky, A. (1977). Features of similarity. Psychological Review, 34(4), 327-352.

Tversky, A. & Kahneman, D. (1974). Judgnent under uncertainty: Heuristics and biases. Science,

 

1_83, 1124-1131.

Tversky, A. & Kahneman, D. (1981). The framing of decisions and the psychology of choice.
§c_ieac_e, 2H (3), 453-458.

Tversky, A. & Kalnneman, D. (1982). J udgnents of and by representativeness. In D. Kahneman, P.
Slovic, & A. Tversky (Eds.) Ju_dgn_ﬂn_t under uncertamtv: Heuristics_ and ﬂags. Cambridge,
UK: Cambridge University Press.

Tversky, A. & Shaﬁr, E. (1992). Choice under conﬂict: The dynamics of deferred decision.

Psychological Science, 3(6), 358-361.

Vaughan, F. E. (1989). Varieties of intuitive experience. In W. H. Agor (Ed) Intuition in

organizations: Leading and managing productively. Newbury Park: Sage Publications.

Wickens, C. D. (1986). The object display: Principles and a review of experimental ﬁndings; (Tech

 

Report CPL 86-6). Charnpaign, IL: Cognitive Psychophysiology Laboratory.
Wickens, C. D. & Andre, A. D. (1990). Proximity compatibility and infomnation display: Eﬂ’ects

of color, space, and objectness on information integration. Human Factors, 2(1), 61-77.

139

Winer, B. J. Single-factor experiments having repeated measures on the same elements.

Woods, D. D., Wise, J. A. & Hanes, L. (1981). An evaluation of nuclear power plant safety
parameter display systems. In Prmeedingaof the Human Factors Socieg 25th Annual
Manning (pp 110-114). Santa Monica, CA: Human Factors Society.

Zey, M. (1992). Decision making: Alternatives to rational choice models. Newbury Park: Sage

Publications.

"lllllllllll'llllf