,l
1 42:5
.

:2" 4
Uza‘lsxm:

-

.-
"3
'p

2 ‘Wt‘ -.
fig’sfil‘t'lgfi

, { vi 1'
23334444441
“ 95%;!

D

V

- 32232
: 4M

I};

w: .
. w- <- _:.. .4 -
.cumgaz...
r". fl.-- -
_ -.
, 332:3...
.
3.11.29
war."
it'u.‘

,-v»4~v
I

: 31'.
4

v.
"i

u~t“"'

I"
:3 “
£33

«Iv—:xr‘n-II

.qu. “V.V

 

 

:r;-

‘3
‘1 $1:

2’

5-"
-0 .

.. -
..
1—3.:

”3.5.3:;-

arvw- "

-...

.N _
..

qu'
..,
4.

'"Wsfiwm’fi X" '
x .
' ‘V

 

..

{v

a

A.
.

We

tr.
‘32

m
1

\' 0
4.5.1160

n-a

. W
15?;
Rb
3,:
4 ‘

.u“

,g- ,,

41“: .
“.4

...J
.4.»»

A
19,3

 

' 5
u‘
0“"

I n
J“

.-o.--

".-4»

.
'1
31~

0.

15,.

.«
"Inc:
«7 Jun... ‘
a. _‘ h“
-..- ~3‘u:'_..-.
u.._ _;; ,
- an... ~,

. '.-‘
. ,,,z;,-...,
4r"

'4?
v
:3“

’. wu’u

 

2""

33:
7-“

7'»
.

. “:11. '43:. n
5319223?“ ‘
'41Sg‘vgt‘qa
1:3”. 4.

‘ ,
. 1 ¥‘.:1N
' . -

.
l I

4.6!: {1:53.3-

11’;v!"\ 4"“ "

_ ‘ fivh‘th’
mu {‘L .
‘0 ’1

_~

I
't

.
‘l
H.
.3
'1

“4?: I

1;“
4“

\n
D

 

3.»

,,

 

, -4. ,
.‘W

-...
”4:31;:‘3

..

‘3
31’11‘

._....

m‘

- ((ir‘fl‘w

.‘w
'.--.._.\- , ‘
m ..r ._ .

hi: 954'.
A 1.: v
‘ .o :'-;'
b ' t‘

.93,“ 5:14
“9? a
' :4
‘ 1‘:

.5
- v-
' "li

Q0-
.4- 1‘:

'1
C V‘ " ‘
,4 3- $3

-v.—

.-

no-
C¢¢-n~' '.
.¢:->.4v~4

1
.e‘u‘.

co.
,4

-..-
.4--
..

H .‘JL'J.

v

""34. i-

5;.
"’2

 

' \
4 3
.9.
file;
't‘f"?‘-§
~ ‘ ‘..| u!
M- :‘M’.
" “$1:
~ . 1.: l y
1 1331:. ‘v‘
:10;
.42

:3
‘l

 

5, .z
I" ‘
in“;
‘ ., ‘

h“

M!”

,2‘::..
.. <

... .

.9

-.-

,—

vfj‘afl‘;
w

"f.
W . :7. 3;:
0.-. . ‘-
arr-:3“?
...
T“

--u
r.

4“
o." -

. ,'.'. ..
‘-
_.—-

'

.c--
,4.-.
" dv"

- 41‘! £1

1} Niki

$44.,

\
l'.‘

lg

v

5-:
...
..
.
.1.
4-5“: -
.
.

L
".15.
'. 4 ,0

,3;
~.«i-

.4311; '
@355
I:

L'l

 

n,- W
313
l 4
‘ n

J
|
I
I

A
.5 h

A

4%”
MA
9‘ .
«z ”2“»!

5.14”
'1
El

:
n

.;

as:
6.1"

I

w
:lfifififiza
"a . '

 

...‘- n,
$35... .-
33’???“

.4.‘
‘5

"£3
«.33.: ‘15"
w." i. Egg?!" 4.—

1.1:
04-"
.

'. n
"3

“ 1‘ :

M4. 8
“431.48%"
’ 'H ' z

.4.
.

IUIWHI’lHlllllHlllllllllllllIll!HIHIIIHIHI

9 31293 01568 7662

J ' LIBRARY I

Michigan State
University

 

 

 

This is to certify that the

dissertation entitled

THE EFFECTS OF INQUIRY METHOD, INFORMATION SHARING AND
INTRAGROUP CONFLICT ON GROUP DECISION MAKING

presented by

Dennis J. Devine

has been accepted towards fulfillment
of the requirements for

Ph.D . degree in Psychology

 

I
/ MMrofessO

Date January 10, 1997

MSU is an Affirmative ActiOn/Equal Opportunity Institution 0—12771

 

 

 

 

 

 

- . ...‘ .

PLACE It RETURN BOX to remove We checkout from your record.
TO AVOID FINES return on or bdore date due.

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU Is An Affirmative Action/Emmi Opportunity Inetitwon
Wm:

 

THE EFFECTS OF INQUIRY METHOD, INFORMATION SHARING AND

INTRAGROUP CONFLICT ON GROUP DECISION MAKING
By

Dennis J. Devine

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

DOCTOR OF PHILOSOPHY

Department of Psychology
1 997

ABSTRACT

THE EFFECTS OF INQUIRY METHOD, INFORMATION SHARING AND

INTRAGROUP CONFLICT ON GROUP DECISION MAKING
By

Dennis J. Devine

Strategic decisions in organizations involve complex, cross-functional issues
which have long-term implications for the viability of the organization, and they usually
involve groups of individuals from various areas of the organization. However, groups
have been shown to be susceptible to process loss, which is viewed in this study as a
failure on the part of group members to share information or adequately utilize
information which is shared. Techniques designed to reduce process loss in group
decision making have foCused on stimulating controversy through the use of inquiry
methods such as Devil's Advocacy (DA) and Dialectical Inquiry (DI). This paper offers
an integrative model of group decision making which focuses on group-level information
processing, and examines the impact of several methods for structuring group discussion
on group process and performance. In general, the inquiry methods used were found to
have little impact on group process or performance but there was some support for the
process relationships in the model. Two group-level variables, cognitive ability and task
knowledge, were the best predictors of group performance. The discussion offers a
revised model and elaborates on future research needed to further specify a model of

group decision making in ill-structured contexts.

 

 

ACKNOWLEDGMENTS

Over the two years spent working on this project, there were times when I felt I
would never finish. I’m sure I’m not the only one. Although the despair and self-
loathing that sets in as the months slip by must be common to those pursuing Ph.D.’s
(oh, I told myselfI wouldn’t do this), the shared misery brings little comfort.
Dissertations remain personal, dirty little guerrilla wars where motivation is ambushed
daily and good intentions are fragged by more hedonistic desires. Still, the war is now
over and, according to MSU, I won. I would like to take the time to thank a few of

those who helped in the war effort, especially those that are still missing.

To begin with, I would like to thank the members of my dissertation committee --
Dan Ilgen, Neal Schmitt, Steve Kozlowski and Rick DeShon (the Joint Chiefs?). They
not only provided a great deal of helpful advice and constructive criticism on this
project but also, during my tour of duty at MSU, provided me with four different role-
models of how to be a good researcher. I learned something important about being an
I/O psychologist from each of them.

I also owe a debt of gratitude to my buddies in the I/O program -- Dave Whitney,

Stan Gully, Jen Hedlund, Eleanor Smith, Matt Smith, Field Marshal Herr William von
Rogers, David “The Mind” Chan, Ken Brown, Dan Weissbein, and Dave

Waldschmidt. These comrades supplied generous quantities of technical advice,

iii

'a. v 4‘ Maximum"

' ‘1
-

 

 

 

sympathetic speculation and emotional support during the long process -- usually, I’m
sure, when they were wishing they had chosen the V8. (Note to self: In the future,
beware of graduate students hanging out in computer rooms looking bewildered.)
It’s hard to imagine doing a dissertation - or even going through graduate school --
without them. Thanks, guys.

Thanks also to Mom, Dad, Cathy and Jim. What good is a Ph.D. without a
family to lord it over? Just kidding, of course (I think).

Finally, I would like to thank my wife for putting up with many late nights, lots of
meaningless rambling, a good deal of distracted listening and a fair amount of
absentmindedness over the two years it took to complete this project. I would also like
to thank her for putting up with the late nights, meaningless rambling, distracted
listening and chronic absentrnindedness. (By the way, I don’t think it’s going to get
any better). N o warning would have been sufficient, but I couldn’t have asked for
more love and support. She even got in on the end of preparing this manuscript and, in
only a short time, learned how comically surreal the formatting process really is.

Thank you, Julie.

iv

TABLE OF CONTENTS

 

 

 

 

List of Tables .................................................................................................................... vii
List of Figures .................................................................................................................. viii
INTRODUCTION .......................................................................... 1
Strategic Decision Making ................................................................................ 2

Strategic Decision Making In Groups ............................................................... 4

Process Loss in Strategic Decision Making ............................................................... 5

A Conceptual Framework for Process Loss ...................................................... 5

Restricted Information Sharing ......................................................................... 8

Poor Information Integration" ..................................................... 16

Summary: Process Loss In Group Decision Making ..................................... 23
Controversy in Strategic Decision Making .............................................................. 25
Controversy-Based Interventions in Individual Decision Making ,,,,,,,,,,,,,,,,,, 28
Controversy-Based Interventions in Group Decision Making ,,,,,,,,,,,,,,,,,,,,,,,, 43

The Forgotten Role of Synthesis ..................................................................... 51
Integration ................................................................................................................ 55

A Process Model of Group Decision Making ................................................. 55

A Prescriptive Model of Group Decision Making .......................................... 61

Summary ......................................................................................................... 66

METHOD ...................................................................................................... 68
Participants ...................................................................................................... 68

Task Overview ................................................................................................ 68

Research Design .............................................................................................. 73

Procedure ................................................................................ 78

Measures ......................................................................................................... 83

Rater Training ................................................................................................. 90

RESULTS ......................................................................................................................... 94

Overview of Results Section ........................................................................... 94
Group Attrition ................................................................................................ 94
Quality of Manipulations and Measures ......................................................... 97
Process Hypotheses ....................................................................................... 109
Inquiry Method Hypotheses .......................................................................... 1 13
Summary ....................................................................................................... 116
EXploratory Analyses .................................................................................... l 18
DISCUSSION ................................................................................................................. 120
Study Contributions ...................................................................................... 120
Study Limitations .......................................................................................... 121
The Process Model ........................................................................................ 126
Dialectical Inquiry in Group Decision Making ............................................. 131
Biased Information Sarnpling and Hidden Profiles ...................................... 133
Future Directions .......................................................................................... 134
Conclusion .................................................................................................... 138
Appendices
APPENDD( A SOUTHEAST AIRLINES, INC.: A BUSINESS
SIMULATION IN THE AIRLINE INDUSTRY .......................................... 140
APPENDD( B Instructions to Groups in the Consensus-Seeking
Condition ...................................................................................................... 172
APPENDIX C Instructions to Groups in the Traditional Dialectical
Inquiry Condition .......................................................................................... 173
APPENDD( D Role Instructions Provided to the Vice-Presidents
Involved in the Dialectical Inquiry Process (TDI & SDI conditions) ,,,,,,,,,,, 174
APPENDIX E Instructions Provided to Groups in the Synthesis
Dialectical Inquiry Condition ........................................................................ 175
APPENDIX F Role Instructions Provided to Vice—Presidents in the
Synthesis Role (SDI condition) ..................................................................... 176
APPENDD( G Information Sharing Check-List Measure ,,,,,,,,,,,,,,,,,,,,,,,, 177
APPENDD( H Observer Rating Form: Intragroup Conflict ,,,,,,,,,,,,,,,,,,,, 187
APPENDIX I Questionnaire Items: Intragroup Conflict ,,,,,,,,,,,,,,,,,,,,,,,, 188
APPENDIX J Observer Rating Form: Process Facilitation ,,,,,,,,,,,,,,,,,,,, 189
APPENDIX K Observer Rating Form: Controversy ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 190
APPENDD( L Observer Rating Form: Controversy (Recoded) ,,,,,,,,,,,,,, 191
APPENDIX M Task Knowledge Measures .............................................. 192
APPENDIX N Questionnaire Items: Implementation Quality ,,,,,,,,,,,,,,,,, 197
LIST OF REFERENCES ................................................................................................ 198

vi

LIST OF TABLES

Table 1 - Empirical Studies of Controversy-Based Decision Aids ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 42

Table 2 - Information Provided to Vice-President Positions in "SouthEast Airlines" ,,,,,,, 71

Table 3 - Role Assignments by Study Condition ............................................................... 77
Table 4 - Sequence of Events in "SouthEast AirlineS" ...................................................... 79
Table 5 - Cell Sizes for Study Conditions and Statistical Analyses ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 96
Table 6 - Interrater Reliability Estimates for Process Measures ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 99
Table 7 - Multivariate Analysis of Potential Confound Variables ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 106

Table 8 - Means, Standard Deviations, Intercorrelations and

Reliabilities for Measured Variables .................................................................... 107
Table 9 - Hierarchical Moderated Regression Results for Hypothesis 1 ,,,,,,,,,,,,,,,,,,,,,,,,, 110
Table 10 - ANOVA Summaries, Hypotheses 6-8 ............................................................ 115
Table 11 - Cell Means and Standard Deviations, Hypotheses 6-8 ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 115

vii

LIST OF FIGURES

Figure 1 - The Moderating Effect of Intragroup Conflict .................................................. 59
Figure 2 - A Process Model of Group Decision Making ................................................... 62
Figure 3 - A Prescriptive Model of Group Decision Making ............................................ 65
Figure 4 - Path Estimates for the Process Model ............................................................. 117
Figure 5 - Information Sharing x Intragroup Conflict Interaction ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 128
Figure 6 - A Revised Model of Group Decision Making ................................................. 130

viii

INTRODUCTION

Groups and teams are ubiquitous in society and increasingly employed by
organizations in the form of production teams, quality circles, committees, advisory
boards, task forces, project groups and semi-autonomous work groups. A strong and
enduring interest in group and team performance has generated a voluminous literature
and a number of reviews (e.g., Cartwright & lander, 1968; Steiner, 1972; Shaw, 1981;
McGrath, 1984; Levine & Moreland, 1990; Bettenhausen, 1991). As a result, we know a
great deal about the factors that determine their effectiveness (Hackrnan, 1990).

Recently, a distinction between "groups" and "teams" has been made by several
reviewers (e.g., Salas, Dickinson, Converse, & Tannenbaum, 1992; Ilgen, Major,
Hollenbeck & Sego, 1993; Saveedra, Barley & Van Dyne, 1993). However, the
distinction has not been honored by most of the literature which has tended to use the
terms interchangeably (Ilgen et al., 1993). As a result, the terms "groups" and "teams" are
used synonymously in this paper to refer to sets of (1) three or more individuals who (2)

interact around a shared goal and (3) are interdependent in some fashion (Steiner, 1986).
Although groups and teams are charged with many different tasks and

responsibilities within organizations, one of their most common fimctions is to make
decisions and solve problems that arise. Typical decision situations involve the division
of work assignments, the sequencing and coordination of activities, purchasing materials,
and allocation of resources. These sorts of decisions are generally encountered on a
regular basis, so organizations develop rules, procedures and policies to standardize and
expedite their resolution (Katz & Kahn, 1976; Galbraith, 1973; Taylor, 1992).

Strategic Decision Mild—11g

Many decision situations of substantial importance to organizations are what some
have called "strategic" in nature. These decisions differ from ordinary decisions in that
they often cannot be forecasted, have important consequences for the entire organization
and involve uncertain relationships between important variables (Taylor, 1992).

According to Taylor (1992), strategic decision making has its roots in the fields of
(1) individual decision making in organizations and (2) business policy and strategic
management. A cursory review of the literature reveals almost as many definitions of
strategic decision making as there are textbooks and papers on the subject (Taylor, 1992).

Mintzbcrg, Raisinghani and Theoret (1976) examined 25 decisions in a variety of
organizations and concluded that strategic decision processes are "characterized by
ambiguity, novelty, complexity and open-endedness" (p. 250). Mason and Mitroff (1981)
characterized strategic decisions as involving many complicated linkages between the
organization and a dynamic and uncertain environment involving ambiguous information
and conflicting goals among interested parties. Shirley (1982) described strategic
decisions as those involving the entire organization and its environment and as ones
which direct and constrain future activities important to the success of the enterprise.
According to Shirley, strategic decisions involve such things as the organizational
mission, goals and objectives, customer mix, product line, geographical service area,
competitive advantage and activities with other organizations. Narayanan and Fahey
(1982) defined strategic decisions as involving important resources or activities for which
there is no precedent or predetermined responses. Thomas (1984) noted that strategic
decisions have little initial structure, long time horizons, political implications, a
sensitivity to environmental dynamics, and affect multiple areas of an organization.
Pearce and Robinson (1988) identified strategic decisions as those made by top

management involving considerable company resources that commit a firm to a course of

3
action with important implications for future profitability and require the coordination of
many functional areas and factors external to the organization. Taylor (1992) stated that
strategic decisions involve novel, ill-structured and complex sets of interdependent
decision problems.

A number of themes emerge from these various definitions. A dominant theme
expressed by most is the importance of strategic decisions to the future well-being of the
organization. Another characteristic is their infrequent, irregular, ofien unpredictable
occurrence. Most reviewers also noted that strategic decisions afiect and pertain to
multiple areas, functions or departments within an organization. Finally, a fourth critical
characteristic separating strategic decisions from other important everyday decisions is
their lack of structure (Ackofl‘, 1974; Mintzberg et al., 1976; Mason & Mitrofl', 1981;
Thomas, 1984; Taylor, 1992).

For the purposes of this paper, a decision situation can be said to be ill-structured
when it involves ". . .decision processes that have not been encountered before in quite
the same form and for which no predetermined and explicit set of ordered responses
exists in the organization" (p. 246, Mintzberg et al., 1976). Moreover, ill-structured
decisions involve uncertainty with regard to the relationships between variables in the
organization and environment as well as interdependency between those variables
(Hirokawa, 1990). As a result, the defining aspect of an ill-structured decision is the
relevance of multiple vieWpoints or perspectives on what needs to be accomplished and
how to go about doing it. For example, before a company decides whether to come up
with a new company logo, it must first decide why a logo is important to company goals,
how much impact the characteristics of the logo will have on those goals, and then
identify the various elements that make a logo "g ."

A primary purpose of this study is to better understand the factors which operate

to afi‘ect group strategic decision making performance. As noted previously, strategic

4

decisions afiect the entire organization, have no precedent, involve multiple
functions/departments and are ill-structured. Typically, multiple perspectives can be
applied in addressing a strategic decision or problem, as well as a wealth of
organizational and environmental data. Because of the above characteristics, "strategic"
decisions are usually made by groups of people rather than individuals (Mintzberg et al.,
1976; Quinn, 1980; Fahey, 1981; Schweiger, Sandberg & Ragan, 1986; Walsh, 1986).
Strategic Decision Making in Groups

In their review of decision making in organizations, Koopman and Pool (1990)
note that "complex decisions are usually not made by individuals, but - after the
necessary preliminary work - by boards of directors, project teams, management teams
and so on" (p.107). They also suggested that the participation of multiple organizational
members from several departments ideally yields a decision that is well-considered and
well-analyzed. However, there is a great deal of anecdotal evidence which suggests
groups often make decisions that are less than optimal - sometimes with disastrous
consequences (Janis, 1972; Janis & Mann, 1977; Ilgen et al., 1993). A number ofcase
studies in the literature have linked a variety of well-known "fiascoes" in world history to
sub-optimal small group decision making, including poor decisions regarding the
appeasement of Hitler at Munich, the escalation of US. involvement in Vietnam, the Bay
of Pigs invasion, the break-in at Watergate, the launch of the space shuttle Challenger, the
attack on the USS Stark, and the shooting down of an unarmed Iranian airbus in the
Persian Gulf. These and other daily events provide a constant reminder of the need (and
room) for improvement in strategic decision making by groups.

Ifwe are to improve group decision making quality, it is necessary to Imderstand
the factors that afiem group performance. The dominant theoretical approach to
\mderstanding group performance was formulated by Steiner over 20 years ago and is

centered around the notion of "process loss." The next section provides an overview of

5
Steiner's (1972) work on group performance and process loss, identifying two specific
kinds of process loss relevant to strategic decision making teams: restricted information

sharing and inadequate utilization of shared information.

Process Loss in Strategic Decision Making

Early research in social psychology painted a rosy picture of the efficacy of
decision making in groups, showing them to be more efiective than individuals (e.g.,
Shaw, 1932). However, using the notion of nominal "statistized" groups, Marquart
(1955) and Lorge and Solomon (1955) showed that groups appeared to perform better
than individuals where each member had an equal and independent probability of solving
the task simply because they were made up of more individuals. Ifthe same number of
individuals working alone was compared to those working together, the advantage of
groups disappeared.

Since that time, a conclusion reached by several reviewers of the sizable literature
on small group performance is that groups do not perform as well as their best individual
member (e.g., Steiner, 1972; Hill, 1982). Although the relationship between individual
performance and group performance is strongly dependent on task type (McGrath, 1984),
across a wide variety of tasks and settings, groups tend to perform better than their
average members but not as well as their best member (Hill, 1982). This suggests that
groups do not combine their resources in an optimal fashion. The decrement between
potential productivity and actual productivity has been termed "process loss” (Steiner,
1972)

A Congeptugl Framework for Process Loss
In 1972, Ivan Steiner conducted a review and re-interpretation of the group

performance literature based on what he called a partial taxonomy of tasks and an

6
input-process-output approach. Steiner differentiated group tasks on the basis of whether
or not group members could divide up tasks so that not all members were working on the
same product. Whereas divisible tasks can be broken down into subtasks and the final
product reassembled by combining the sub-products of the various group members,
unitary group tasks involve a single group product that cannot be distributed among
members and recombined. Steiner went on to assert that task type is a primary
determinant of how individual inputs are combined to form a unitary group product. For
unitary tasks, the group product is essentially a function of some selected individual in the
group. According to Steiner, the function relating individual performance to group
performance is inherently different across task type. Steiner focused his analysis on
unitary tasks, identifying four types of such tasks and corresponding "rules" regarding
which individual within the group would provide the group's final product: disjunctive,
conjtmctive, additive and discretionary tasks.

Disjunctive tasks allow the group to utilize the efl‘orts of their best member as the
group product (e.g., spelling a dificult word, answering a history question). In
conjunctive tasks, the nature of the task is such that group performance becomes a
function of the worst member of the group (e.g., time taken for a group to cross a river).
Additive tasks utilize the inputs of every group member, resulting in group performance
becoming a function of the average member. Finally, discretionary tasks allow the group
freedom to determine how individual inputs will be transformed into a group product -
according to disjunctive, conjunctive or additive rules.

A primary contribution made by Steiner was his formalization of the manner in
which groups inputs are converted to outputs. According to Steiner, the actual observed
productivity of a group is equal to its potential productivity minus process loss. "Process
loss" is viewed as the additive result of two phenomena - coordination loss and

motivation loss.

7

Coordination loss refers to a decrement in process that arises from the need to
coordinate the efforts of multiple persons, for example, when a group of people
participating in a tug-of-war must synchronize their tugs. In contrast, motivation loss
refers to an unwillingness on the part of individuals to contribute their efforts to the group
product.

Since Steiner's work on the notion of process loss," a great deal of research has
shown that groups can suffer from process loss in several ways. For example, Latane and
his colleagues have extensively studied the "social loafing" effect (Latane et al., 1979;
Latane, 1986). Social loafing is a type of motivation loss represented by a decreasing
output/person ratio as group size increases. Social loafing appears to be a robust
phenomenon, having been documented over a variety of physical tasks (i.e., clapping,
shouting, rope pulling) and even cultures.

Motivation loss and coordination loss seem well-suited to explaining process loss
in groups performing tasks with some physical component. However, more relevant to
group decision making settings is a conceptual fi-amework ofi‘ered by Larson and
Christensen (1993) which identifies six general functions that occur in group problem
solving that potentially involve process loss: problem identification, problem
conceptualization, acquiring information, storing information and retrieving information.

They point out that, with respect to groups problem solving, process loss is conceptually
equivalent to inefficiency in mechanisms by which information known to individuals in
the group is converted into a knowledge structure characteristic of the group as a whole.

Problem identification is seen as the collective recognition by all members of the
group that a problem exists which must be addressed. The process by which group
members come to agree on the way a perceived problem is categorized is problem
conceptualization. The third fimction, acquiring information, is the process by which
groups identify and attain information perceived to be task-relevant. Alter individuals

8
acquire information, it must typically be stored in some fashion before it can be shared
with other group members. Once in the group setting, information must then be retrieved
before it can be shared with other members. Finally, after information has been shared
with other group members, it must be integrated and implemented in a collective
decision.

Larson and Christensen (1993) note that little work on groups has been conducted
focusing on how individual processes are related to group-level processes and outcomes.
Their "information retrieval" function is an exception to this rule. "Information retrieval,"
or essentially the sharing of individual information with the entire group, has previously
been identified and studied from two different theoretical perspectives. The next section
reviews research conducted within these two paradigms that examines defects in group
decision making that resulted from inadequate information sharing.

Restricted Information Sharing

Schweiger, Sandberg and Ragan ( 1986) note that group processes do not ensure
that managers will adequately explore available information necessary for making good
decisions. One way that this can happen is information known to individuals is not made
available to the group as a whole. Essentially, if individuals in the group are unwilling or
Imable to share relevant information not known to other members, that information
cannot be used by other group members. Restricted information sharing between
individuals and other group members has been studied from two vantage points in the
literature: the "groupthink" syndrome (Janis, 1972; Janis & Mann, 1978; Janis, 1982) and
the "biased information sampling" efl‘ect (Stasser & Titus, 1985; Stasser & Titus, 1987).
Groupthink is thought to result from an overriding desire to preserve positive relations
among group members, which leads to a restriction in the amount and type of information
considered by the group. In the case of biased information sampling, groups center their

discussion around information that group members hold in common. Both of these types

9
of process loss result in the failure on the part of one or more group members to share
potentially valuable information. The next two sections review the literature with respect
to these two phenomena.

"Groupthink." Groupthink is a term that seemingly needs little introduction. In a
1972 seminal work entitled "Victims of Groupthink," Irving Janis coined the term
"groupthink" and defined it as "a mode of thinking that people engage in when they are
deeply involved in a cohesive in-group, when the members‘ strivings for unanimity
overrides their motivation to realistically appraise alternative courses of action" (p. 9).
Taylor (1992) defined groupthink as "a collective pattern of defensive bolstering used by
a decision making group to shield itself from negative information and criticism" (p.
980). In general, the term has been used to explain ". . . why people in authority
frequently act contrary to enlightened self—interest by making decisions that are likely to
be counterproductive" (Whyte, 1989).

Janis and his colleagues (Janis, 1972; Janis & Mann, 1977; Janis, 1982) have
identified a model of groupthink little changed in over 20 years. Janis (1972) identified
five antecedents of groupthink, eight "symptoms" of groupthink and seven defective
decision making processes. In essence, the theory suggests that high levels of the various
antecedents lead to the exhibition of a number of groupthink symptoms and the use of
defective decision processes. Groups sufl‘ering from groupthink accordingly make
decisions which have a low probability of successful outcome.

Janis classified the five antecedents of groupthink into three primary categories:
group cohesion, structural faults of the group and a provocative situational context.
Group cohesion referred to the level of interpersonal attraction and the extent to which
the group was "close-101i ." Included in the category of "structural faul " were factors
such as insulation from the outside world, lack of an impartial leader, lack of procedural

norms and excessive member homogeneity. Among the elements of a provocative

10
situational context, Janis identified two focal aspects, high stress and low group "self-
esteem." Stress was thought to result from perceived external threat and low hope of
finding a better alternative than the existing plan. Low group self-esteem was thought to
result from inadequate efforts to resolve the problem in the recent past.

In essence, when levels of group cohesion and situational stress are high and
structural aspects of the group are such that there is pressure to suppress individual
dissent and come to a quick resolution, groupthink is hypothesized to manifest itself
through a variety of "symptoms." These symptoms of groupthink include the illusion of
invulnerability, moral certainty, collective group rationalization, stereotyped conceptions
of the "opponent," self-censorship within the group, the illusion of unanimity, pressure on
dissenters and the use of "mindguards" to ward ofl‘ anxiety.

In terms of actual decision making behaviors, Janis (1977, 1982) noted seven
process characteristics associated with groupthink: an incomplete identification of
alternatives and objectives, a failure to re-examine the preferred choice or rejected
alternatives, poor information search, biased and selective information consideration and
a failure to develop contingency plans.

Despite the popularity of groupthink as an explanation of how limited information
processing leads to poor group decisions, relatively little empirical research has addressed
the groupthink phenomenon. Mullen and Copper (1994) counted more reviews of the
literature than empirical studies. In general, the research on groupthink can be divided
into two categories: retrospective analyses of decision making episodes and laboratory
studies.

Several early studies of groupthink utilized a case study approach to examine the
decision processes behind major policy decisions in history (e.g., Janis, 1972; Janis &
Mann, 1977; Tetlock, 1979; Janis, 1982; Smith, 1984; Hensley & Griflin, 1986; Esser &
Lindoerfer, 1989) In a comprehensive review of the groupthink literature, Aldag and

1 1
Fuller (1993) identified six retrospective case studies and nine empirical studies
pertaining to groupthink. In general, these post-hoc analyses studies have tended to find
support for the groupthink model (Aldag & Fuller, 1993).

In contrast, studies of the groupthink phenomenon conducted in laboratory
settings have been less supportive (Aldag & Fuller, 1993; Park, 1989). A number of
studies have attempted to manipulate levels of group cohesion along with other structural
or situational variables such as leader style/behavior (Flowers, 1977; Fodor & Smith,
1982; Leann, 1985), the desirability of conflict (Courtright, 1978), and task type
(Calloway & Esser, 1984). For the most part, these studies have used groups of
undergraduates and tasks involving hypothetical business scenarios or complex
management simulations (although see Calloway & Esser, 1984 for a notable exception).
A recent meta-analysis examining group cohesion and the quality of decision making
found that high levels of group cohesion are necessary but not sufficient to produce
characteristics of groupthink (Mullen et al., 1994).

Of particular interest to this study are two experiments which examined the effect
of groupthink antecedents on information sharing among group members. Fodor and
Smith (1982) examined the effects of group cohesion and leader "need for power" on
several process and outcome measures for groups of undergraduates solving a business
scenario. Information relevant to the task was divided and distributed among group
members, with a "self-censorship" variable created to measure the number of information
cues group members shared from their respective "role sheets" during discussion. One
experimenter observed each group during the study and coded the number of facts
introduced by each member during discussion. The combined number of facts introduced
by groups was found to be strongly and positively related to the number of alternatives
considered (r = .71). Although Fodor and Smith found that groups introduced more
factual information into discussion when group leaders had a low need for power, group

12
cohesion was not found to have a significant effect on the number of facts introduced into
discussion by group members.

Another study by Leana (1985) also examined the eflect of group cohesion and
leader behavior on group processes and outcomes. Groups of undergraduates were given
the task of solving a hypothetical business problem and with each individual assigned to a
specialized role with access to unique task-relevant information. Experimental sessions
were tape recorded and subsequently coded for the number of facts introduced from the
player role sheets. Unlike F odor and Smith (1982), a leader characteristic (leadership
style) did not have an efi'ect on the number of facts introduced but group cohesion was
found to be related to information sharing. However, contrary to the groupthink model,
cohesive groups shared greater than 50% more information than noncohesive groups.

SMIL- The groupthink model has maintained its appeal in spite of operational
dimculties in testing it and limited empirical support (Moorhead, 1982; Tetlock et al.,
1992; Park, 1990; Aldag & Fuller, 1993). Retrospective case studies have tended to
support the central tenets of groupthink whereas laboratory studies which have focused on
testing the linkages between the proposed causal antecedents and the symptoms of
defective decision making have not (Aldag & Fuller, 1993).

In keeping with the general lack of empirical attention to the overall groupthink
model, few studies have directly addressed the role of information sharing (i.e., "self-
censorship" in groupthink terminology) in group decision making. Two studies that did
look at information sharing (i.e., Fodor & Smith, 1982; Leana, 1985) came to essentially
Opposite conclusions about the antecedent factors that lead to low information sharing
(cohesion as opposed to leader style), but neither study explicitly measured the
relationship between self-censorship and group decision making performance. Thus,
despite its central role as a prOcess defect in J anis' model, the role of information sharing

is unclear. The current study will address this omission by directly measuring

13
information sharing and assessing its relationship with group performance. In the next
section, we review literature on a second type of process loss associated with reduced
information sharing -- biased information sampling.

Biased information sampling. A common technique used by groups to exchange
information, resolve differences and make decisions is face-to-face discussion among
group members (Schweiger et al., 1986). Recent theoretical developments concerning
how information is sampled by groups during discussion has led to a formal model of
information sampling by groups (Stasser & Titus, 1985).

In essence, information sampling theory predicts information is sampled on a
probabilistic basis by groups during group discussion. An item of information is
"sampled" by a group when an individual with access to the information one shares that
item with the group as a whole. The probability that a given piece of information will be
sampled is a ftmction of the number of people who have access to that item. Assuming a
constant probability associated with any given group member recalling and sharing any
particular piece of information he or she possesses, the chances of mentioning a given
piece of information increase as a fimction of (a) the number of individuals within the
group that have access to the information before group discussion and (b) as the size of
the group increases.

A significant and unfortunate implication of information sampling theory is that
group discussion tends to focus on shared information known to all or most group
members. In other words, potentially important Imshared information is less likely to be
mentioned to the group and therefore remains "unusable." According to the theory, the
bias in favor of sampling previously-shared information increases with group size. The
predictions of information sampling theory with respect to biased information sampling
have been supported across a number of empirical studies (Stasser & Titus, 1985; Stasser
& Titus, 1987; Stasser, Taylor & Hanna, 1989; Stasser & Stewart, 1992).

14

Stasser and Titus (1985) constructed a task in which three hypothetical
"candidates" were running for student body president and asked group members to decide
which candidate was most suitable for the job. When all information for each of the three
candidates was available to all group members, 83% of the groups preferred Candidate A.

When the available information on each candidate was divided and given to different
group members, only 18% of the groups chose Candidate A. Stasser and Titus refer to
this as a "hidden profile." Since individuals in the unshared information conditions
collectively possessed all available information, the most likely explanation for the
difference in preference rates is a failure to share needed information. Free recall data
gathered before and after group discussion indicated that group members "learned" very
little unshared information during the course of group discussion. Stasser and Titus
(1987) replicated this finding and found that substantial gains in terms of learning
tmshared information were made only when there were very low levels of unshared
information in the group prior to discussion. However, neither study directly measured
information sharing by groups, nor was it logically possible to determine the detriment of
restricted information sharing on group performance since there was no way to identify
which alternative was best for the job.

Stasser et al. (1989) extended the two earlier studies by directly measuring
information sharing during group discussion. Using the same candidate selection task
and audio recordings of group discussion, they found that, across all conditions, group
discussions included 46% of the information cues shared by one or more group members
but only 18% of the unshared information cues. As predicted by the model, the sampling
bias in favor of shared information was greater in larger groups (six-person versus three-
person groups) and when a higher percentage of information was known to multiple
group members before discussion (66% versus 33%). In addition, shared information

was significantly more likely than Imshared information to be repeated in group

15
discussion, resulting in only about 5% of unshared information considered and then
reconsidered during discussion. Furthermore, Stasser et al. found that specific
instructions to examine all available information before making a decision led to an even
stronger bias in favor of previously shared information.

Stasser and Stewart (1992) followed up these studies by investigating the effects
of task type. In this study, groups of students were asked to read a murder mystery after
being led to believe that they either did (solve set) or did not (judge set) have suficient
information to solve the crime. When critical cues were unshared before group
discussion, only 35% of the judge set groups identified the correct murderer whereas 67%
of the solve set groups were able to identify the proper culprit.

The findings of Stasser and his colleagues suggest that group discussion is not a
good device for disseminating unshared information. As Stasser et al. (1989) noted,
groups tend to focus their discussion on what is already known, avoiding the
consideration of potentially relevant unshared information. With respect to ill-structured
strategic decisions, it may be the case that groups do not optimally use idiosyncratic
information held by individual "experts." Given that strategic decision making groups are
composed of diverse "experts" precisely because they are expected to share unique
information and assumptions, it may be that strategic decision making groups do not
utilize their primary potential advantage.

Summa_ry: Restricted Information Sharing. Past research and theory on group
decision making has identified two mechanisms leading to process loss in the form of
failure to exchange necessary task-relevant information: groupthink and biased
information sampling. Both theories share a cognitive focus centered around two
assumptions: (1) the quality of group decision making is a positive function of the
amount of information known by all group members and (2) individuals bring difl‘erent

informational resources to the group decision making context.

16

Although little research has directly addressed the role of restricted information
sharing in contributing to poor group decisions, both models highlight the importance of
information sharing among group members. Those studies that did directly measure
information sharing among group members were not in a position to relate it to decision
making quality (Fodor & Smith, 1982; Leana, 1985; Stasser et al., 1989). Stasser and his
colleagues have found that groups often do not end up discussing information known only
to one or few group members.

To this point, the discussion has focused on a type of process loss that arises when
group members fail to share information necessary for optimal decision making.
However, in order for relevant information to translate into good decisions, that
information must be effectively utilized once it has been shared. A second type of
process loss in group decision making occurs when group members share critical
information but disagree on the manner in which the information should be used. Such
disagecment may lead to conflict among group members which prevents the optimal
integration of information. In the next section, we review literature on the role of
intragroup conflict in inhibiting the integration of information provided by individual
group members.

Poor Information Integr_ation

One of the six processes noted by Larson and Christensen (1993) as necessarily
occurring in group problem solving is the integration and combination of information
available to the group. Although there are a number of factors that may lead groups to
process information in a less than optimal fashion, we review evidence in this section
suggesting that one of the most common of group phenomena, conflict among group
members, may lead to sub-optimal integration of shared information. Experimental
social psychology has a long tradition of interest in conflict, and a number of research

paradigns (e.g., Prisoner's Dilemma, Resource Conservation, Coalition Formation) have

1 7
arisen around the study of particular types of intragroup conflict (Argote & McGrath,
1993). Unfortunately, little empirical research has addressed the effects of intragroup
conflict on group decision making.

The Nature of Conflict. Conflict is so ubiquitous that it is considered by many to
be a ftmdamental characteristic of human interaction (Pruitt & Rubin, 1986; Patton et al.,
1991). Pruitt and Rubin (1986) have called attention to the fact that several of the most
influential theorists of the twentieth century (e.g., Darwin, Marx, Freud) have based their
work around the notion of conflict. However, at the same time, they note the term has
been used so broadly and in so many difierent contexts it is in danger of losing its status
as a singular concept.

What exactly is “conflict?” Tillett (1991) provided the following overview:

Conflict is an inevitable and pervasive aspect of human life. . . . It arises within

individuals and between individuals. It takes place within and between groups,

organizations, communities and nations. Conflict occurs at home and at work and

in the neighborhood. . . . Conflict is sometimes violent, but more often not. . . .

Much of it exists within the mind and is expressed in words . . . . Conflict is

popularly equated with fighting, and is generally seen as destructive, unpleasant
and undesirable. It is usually suppressed, avoided, concealed or fought over . . ."

(p. l).

The American College Dictionary defines conflict as “a state of disharmony
between incompatible or antithetical persons, ideas, or interests; a clash.” In a similar
fashion, Webster’s Dictionary defines conflict as “a sharp disagreement or opposition, as
of interests, ideas, etc.” Conflict is perceived as synonymous with a number of other
words as well, including destruction, anger, disag'eement, hostility, war, anxiety, tension,
alienation, violence and competition, dissension, strife, friction, disagreement, dispute,
quarrel, war and fight (Frost & Wilmot, 1978; Tillett, 1991). As such, it is clear that
conflict in a popular sense is equated with negative emotions and hostile behavior.

Further, Tillett (1991) has pointed out that conflict can occur at difl’erent levels of
society. Interpersonal conflict exists between two or more people, while intragroup

l 8
conflict exists within a group and involves two or more subgroups within the larger group
where subgroups may be as small as a single individual. Intergroup conflict, or conflict
between groups, has perhaps received the most attention in the literature. Finally, conflict
can also occur at the international level between two or more countries. Although it
seems reasonable to believe that conflict has a number of basic attributes common to any
level, it is quite likely that the antecedents and consequences of conflict differ according
to level.

Given conflict's many connotations, it is not surprising that researchers have
defined the term in different ways. Adding to the confusion, some researchers have
defined conflict in general terms while others have focused on conflict at a particular
level. Coser (1967) defined conflict as "a struggle over values and claims to scarce status,
power and resources in which aims of the opponent are to neutralize, injure or eliminate
the rivals" (p. 8). Similar to Coser's definition, Deutsch (1973) ofl‘ered a widely-adopted
definition of conflict that focused on “incompatible activities” between two or more
parties, where “incompatible activities” were seen as those that prevent, obstruct,
interfere with, injure or make the actions of another party less effective. Mack and
Snyder (1973) noted that conflict includes two or more interacting parties, positional or
resource scarcity, attempts to acquire or exercise power, and behaviors intended to injure,
thwart and control others. Frost and Wilmot (1978) defined conflict as "an expressed
struggle between at least two interdependent parties, who perceive incompatible goals,
scarce rewards and interference from the other party in achieving their goals" (p. 9).

Pruitt and Rubin (1986) defined conflict as a “perceived divergence of interest, or a belief
that the parties’ current aspirations cannot be achieved simultaneously” (p. 4). Burton
(1988) described conflict as “a relationship in which each party perceives the other’s
goals, values, interests or behaviour as antithetical to its own” (p. 11). Tillett (1991)

defined conflict as existing “when two or more parties perceive that their values or needs

19
are incompatible” (p. 7). Finally, Zander (1994) noted that a situation involves conflict
“if two or more parties disagree over what the other ought or ought not to do -- when each
side knows what should be said or done and knows that opponents’ views are wrong” (p.
112).

There are a number of common elements that stand out in these definitions:
perceived diflerences in values, needs or goals; scarce resources; and incompatible or
hostile behaviors. The various definitions reviewed above suggest that conflict is a
complex construct characterized by phenomena on three general dimensions: affective,
behavioral and cointive. Conflict involves strong negative emotions which are
experienced as distinctly tmpleasant by individuals in the g'oup. These negative affective
reactions include tension, frustration, anger, fear, hopelessness, wormded pride and even
depression (T josvold, 1985; Pruitt & Rubin, 1986; Bettenhausen, 1991; Tillett, 1991).
Tillett (1991) noted that the presence of strong negative emotions as something that
distinguishes conflict from more mundane problems and disputes. With respect to
behaviors, conflict clearly involves a set of negative interpersonal behaviors on the part of
one or more group members that can be interpreted as hostile or degrading by other
members (Coser, 1967; Deutsch, 1973; Frost & Wilmot, 1978; Zander, 1994). Finally,
with respect to cogtition, conflict is associated with distortions in the thought patterns of
group members. In particular, conflict may invoke irrational cognitive processes that
result in the use of over-simplification, exaggeration, extreme generalization, defense
mechanisms, rigid and inflexible adherence to one’s belief (Frost & Wilmot, 1978; Pruitt
& Rubin, 1986).

It is important to recognize, however, that these definitions represent attempts to
define conflict in general terms as opposed to conflict at any particular “level” identified
by Tillett (1991). While thereis certainly some value in this, it is also possible to define

conflict in a more level-specific fashion that retains the essential elements of the construct

20
while introducing aspects of conflict that may be unique to a particular context (i.e.,
level).

Conflict Within Grogpg. Combining the basic features of conflict and these more
observable features of conflict within g‘oups, intragroup conflict can be defined as a
group-level process involving: (1) negative affect, (2) hostile/degrading interpersonal
behaviors, and/or (3) irrational and/or non-task-related cointive processing on the part of
two or more goup members as a result of interpersonal interaction. It is assumed that
conflict among two or more group members can affect the entire group. In general,
intrag'oup conflict can inhibit group-level processing of information by directly
absorbing group resources in the form of collective discussion time and individual
cognitive resources. For instance, individuals arguing about what the group should do
may take up an excessive amount of a group’s available “air time,” preventing
constructive discussion of other issues. Arguments and overt conflicts may also serve as
a distraction, focusing group members’ attention on their own reactions or on ways the
conflict can be reduced, circumvented or even heightened. As a result, intragroup conflict
does not have to involve all group members (or for that matter, even be perceived by all
group members) to afl’ect g'oup-level information processing.

Conflict can arise within groups or organizations in any number of ways. Patton
et a1. (1991) stated that conflict can arise from almost any change, noting “Any perceived
changes, ranging from leadership roles to group structure to activities to new
membership, may provoke conflict. The conflicts are inevitable; the nature of the group
will determine whether they are handled openly or reduced to the level of a hidden
agenda” (pp. 119-120). Zander (1994) noted that conflicts within organizations can arise
from differences over such things as the unwillingness of members to accept leaders,
budget allocations, who should speak for the unit, how lawsuits should be settled,

strategic planning, and the relative attention paid to various departments. Furthermore,

2 1
Zander identified a number of conditions that increase the potential for conflict, including
a lack of systematic procedures for the group, scarce resources, multiple plans or goals for
the group among group members, and confrontational procedures such as devil’s
advocacy.

Conflict within groups develops through various stages or steps in a sequence.
Tillett (1991) noted that intragroup conflict is characterized by an ongoing sequence of
problems or disputes over seemingly minor points or issues that bring on irrational,
emotional or extreme reactions on the part of involved group members, as well as
excessive discussion or argument. Zander (1994) described the following sequence in the
development of intragroup conflict: (1) Members recoinze they disagree, (2) They
confront/attempt to persuade one another, (3) Initial positions “harden,” (4) Rationality in
thinking and communication decreases, (5) Members show hostility, (6) Coercion and
imitation tactics make their appearance, (7) Language becomes “stronger,” (8) Members
begin to feel and exhibit a lack of trust towards other members and, finally, (9) Emotional
waning occurs as members become fatigued.

Similar to the pattern identified by Zander (1994), Pruitt and Rubin (1986)
identified a number of general transformations take place within groups experiencing
conflict, including “Light -> Heavy,” “Small -> Large,” “Specific -> General,” “Doing
Well -> Winning -> Hurting Opponents,” and “Few --> Many.” “Light to Heavy” refers
to the tendency for group members to use increasingly direct, overt influence tactics.
“Small to Large” addresses the tendency for issues to proliferate once initial differences
are discovered. “Specific to General” corresponds to the general trend for issues to
increase in scope, while “Doing Well to Winning to Hurting” characterizes the shift in
group member’s goals from those that focus on group well-being to a relatively narrow
vindictiveness. Finally, “Few to Many” represents the tendency for conflicts to draw in

neutral “observers” and so increase in size.

22

One of the fundamental dynamics of groups experiencing conflict is a heightened
sensitivity to the possible loss of “face” (Pruitt & Rubin, 1986; Tillett, 1991). Pruitt and
Rubin (1986) describe how changes take place within the group centered around this
dynamic:

Zero-sum thinking develops - it’s either victory for them or victory for us. New
goals come to the fore: to look better than, punish, discredit, defeat, or even destroy the
adversary. The capacity for empathy with the adversary is eroded. There are also
changes in the approach taken to group decision making: Positions become rigid, there is
little room for compromise, and there is a dearth of imagination and creativity. Emphasis
is placed on proving how tough and unyielding one is, so as to persuade the adversary that
one cannot be pushed around” (p. 93).

Although the efi‘ects of conflict on individuals are relatively well-known, the
effects of conflict on group processes and outcomes have received little research attention
(Zander, 1994). At an individual level, conflict involves feelings of agitation, annoyance,
and frustration; high levels of interpersonal hostility; and distorted cognitive processes
(Tjosvold, 1985; Tillett, 1991). Further, research has shown that conflict leads to
psychological withdrawal, low commitment to group goals and decisions, low satisfaction
with group process, and reduced preference to work with the same group members in the
future (Levine & Moreland, 1990; Bettenhasuen, 1991). While little research has
addressed the issue, a number of researchers have speculated that conflict is harmful to a
variety of g'oup-level outcomes.

In particular, it seems likely that intrag'oup conflict will have a negative impact
on the ability of group members to optimally utilize information shared by members and
thus available to the group. The previous review has described a number of mechanisms
that may result in a failure to adequately use shared information as a result of diverted or
wasted group resources. Furthermore, a primary dynamic underlying the sub-optimal
utilization of group resources is a fear of losing “face.” As members present their views

and get rebufl‘ed, they become angy, frustrated and impatient. As they do so, they are

less likely to maintain an open mind towards the input of others and more likely to rigidly

23
advocate their own viewpoint and corresponding plan. In such a case, the process by
which groups "build" a constructive compromise alternative may be reduced to a subset
of vocal or high-status group members and the information salient to them (Walsh et al.,
1988). Thus, the construction of a final "group" plan may be largely a product of political
factors and based on a reduced subset of the information offered by group members
during group discussion. Decision quality will likely suffer as a result.
Summary: Process Loss in Group Decision Mak_in_g

Schweiger et a1. (1986) note that "the nature of interactions and the processes by
which information is shared and evaluated appear to be critical factors in the efi‘ective
strategic decisions by top management" (p. 52). An assumption shared by the work of
Janis (1972; 1982) and Stasser and his colleagues (Stasser & Titus, 1985; Stasser & Titus,
1987; Stasser et al., 1989; Stasser & Stewart, 1992) is that greater information sharing by
individuals with access to information the rest of the group does not have leads to better
group decisions. The preceding discussion has identified two potential types of process
loss involving information available to strategic decision making groups: (1) a failure on
the part of individuals to share relevant information and (2) a failure to adequately
integrate information which is shared.

With regard to inadequate information sharing, research on groupthink and
information sampling theory has identified a variety of characteristics (e.g., time and
conformity pressures, high levels of cohesion, external threat, highly distributed crucial
information) that may impair the ability of groups to "draw out" information held by a
subset of members so that it can be utilized at a gem level by all members. Ideally, all
information known to individual group members will be made available to all other
members during discussion. Theory and research on g'oupthink and biased information
sampling suggest that this rarely occurs.

24

A second type of process loss that may occur in strategic decision making groups
is the failure to adequately utilize information which has been shared by individuals in
forming a final group alternative. One of the factors that may lead groups to reduce their
information processing and base their collective decisions around only a small subset of
available information is the presence of conflict among members. Conflict is endemic to
groups and results in negative afl‘ect, hostile behaviors and distorted cognition at the
individual level. The efl‘ect is to reduce or divert group-level information processing
resources from the task at hand. The work of Walsh and his colleagues (Walsh, 1986;
Walsh & Fahey, 1986; Walsh & Henderson, 1988) suggests that, in such situations,
groups will use the problem conceptualizations and supporting information of those
group members which are perceived as most dominant, powerful or expert. This in tum
may lead to a failure to integ'ate all available information in coming to a final group
decision and, as a result, lower g'oup performance.

In spite of the intuitive and analytical reasons for avoiding conflict in group
decision making, a number of researchers have noted positive aspects of intragroup
conflict, including the increased exchange of information among group members and
expansion of the alternatives under consideration (Frost & Wilmot, 1978; Pruitt & Rubin,
1986; Patton et al., 1991). While recognizing these benefits, other writers have
associated these benefits with different labels, including "cognitive conflict" (Priem &
Price, 1991; Amason, 1995) and "controversy" (T josvold, 1985). The proliferation in
terms has led to confusion in the literature. Indeed, a popular approach to increasing the
efi‘ectiveness of group decision making involves the introduction of "conflict" into goup
discussion (Schweiger, 1990). Given the confusion already surrounding "conflict" and
the theoretically-meaningful distinction which can be made between positive and negative

aspects of the term, this paper'will differentiate "conflic " and "controversy." The next

25
section reviews the literature on controversy and controversy-based interventions in group

decision making processes.

Controversy in Strategic Decision Making

Mason (1969) noted that strategic decisions are often made by top management
groups after turning a problem or issue over to a technical or functional "expert" for
consultation. Subsequent recommendations made by such experts to management
decision making groups are typically based on numerous simplifying assumptions that
usually go unrecognized and, therefore, unchallenged. Given the wealth of data that
apply and the validity of multiple perspectives in any ill-structured situation, Mason
suggested that routine implementation of these unilateral recommendations generally
leads to poor decisions. In order to avoid the pitfalls associated with this approach,
Mason (1969) and Mason and Mitrofl' (1981) have advocated the use of techniques that
stimulate controversy among group members.

Much of the research that has been done on the nature of controversy has been
conducted by Dean Tjosvold and his colleagues. Tjosvold (1985) offers the following
definition of controversy:

Controversy is a special kind of conflict and occurs when one person’s ideas,

opinions, conclusions, theories and information are incompatible with another’s

when they discuss problems and make decisions. . . . Controversy involves
difiemnces of opinion that at least temporarily prevent, delay, or interfere with
reaching a decision. Persons in controversy have opposing views about how they
should proceed, and face the pressure to resolve these difi‘erences in order to reach

a decision and move forwar ” (p. 22).

At an operational level, controversy can be seen as the extent to which group
members (1) express doubts about the efficacy of an existing plan or (2) identify multiple
perspectives or plans for achieving a group's primary goal. Elaborating on this definition,

there are at least three types of behaviors that indicate controversy among group

26
members: (1) Indirect challenge, (2) Explicit disagreement, (3) Presentation of opposing
viewpoints. Indirect challenge refers to instances where group members express only
partial agreement for a position or recommendation, present contradictory information
without open disag'eement, or ask questions that indicate skepticism or doubt concerning
a particular point of view (e.g., "Why?"). Explicit disagreement indicates a lack of
support for a plan or position that has been taken, and implies that the speaker believes
there is another, as-yet-unidentified way to do things. Finally, controversy is most
directly evident when group members identify, discuss and debate alternative plans or
viewpoints. It is useful to distinguish these dimensions of controversy in that they all
reflect the same underlying phenomenon yet may not occur or manifest themselves with
equal frequency.

At this point, is it useful to explicitly distinguish intragroup conflict and
controversy. As noted previously, a defining characteristic of conflict is negative afiect
among g'oup members that is interpersonal in nature. On the other hand, controversy
involves disagreement among group members without the negative or interpersonal
aspects. The bases of conflict are personal and emotional; the bases of controversy are
ideological and cognitive in nature. Conflict may be caused by factors completely
unrelated to the g'oup's position or goal, for example past interactions between group
members, whereas controversy is task-related and need not involve any conflict at all,
such as when two friends question one another while "agreeing to disagree." As noted
previously, the occurrence or exacerbation of existing conflict may cause group members
to use information sub-optimally. However, controversy as defined by Tj osvold should
lead to a broadening of ideas and alternatives considered by the group, and may also
cause group members to share information as they are called upon to explicate their
difl'erent positions. As such, while conflict initiates mechanisms which may impair group

functioning, controversy should positively impact g'oup functioning by aiding in the

27
identification of alternative plans and promoting information sharing among group
members.

The assumption underlying controversy-based techniques is that group decision
making performance can be improved by requiring group members to identify and
critically examine the assumptions underlying their ideas, thus "rooting out" diflerences
that might otherwise go unnoticed. Assumptions which survive such scrutiny are
hypothesized to be more likely to be valid than those that do not (Mason, 1969; Janis,
1972; Mason & Mitrofl‘, 1981). Controversy is thus intended to serve as a "perspective-
broadening" tool helping to ameliorate the process losses associated with group decision
making. Unfortunately, little research has addressed the efl‘ects of conflict in the type of
setting most likely to occur in actual organizations - situations in which small gems of
functional "experts" share many of the same goals (e. g., financial profit) yet disag'ee on
how to implement plans to achieve those goals (McGrath, 1984; Levine & Moreland,
1990).

Tjosvold (1985) notes that a number of factors influence the level of controversy
in a group, including membership competition, forming subgroups, leadership style,
openness norms and decision making rules. Specifically, Tjosvold states that “Forming
subg'oups that are assigied opposing positions on the issue in question is a direct way to
structure controversy” (p. 34). A number of studies have now attempted to examine the
eflicacy of introducing controversy into the minds of decision makers in organizational
settings. In particular, Dialectical Inquiry and Devil’s Advocacy approaches assmne that
controversy will improve understanding of underlying issues and lead to the creation of
efi‘ective strategies (T josvold, 1985). However, Tjosvold goes on to note that “these
approaches have neither specified the interpersonal contexts and processes that facilitate
the debate, nor incorporated research findings on the dynamics and outcomes of

controversy” (p. 33).

28

Tjosvold (1985) claims that “Strategic decisions typically evoke controversy
because they are complex and involve persons from different groups and departments
within the organization who evaluate proposals from a variety of perspectives” (p. 33).
Given that strategic decision settings provide fertile ground for controversy, it is
important to determine the degree to which controversy affects group processes and/or
outcomes of interest. Although not explicitly intended to increase information sharing
among g'oup members, it seems likely that the process of challenging and debating
assumptions will also produce increased information sharing among group members. In
essence, when group members are faced with the question, "Well, why do you think
that?," they will, at some point, share the data on which their beliefs are founded. As a
result, controversy should yield greater information sharing among group members as a
by-product of the assumption-challenging procedure. At the same time, controversy also
seems likely to stir intragroup conflict between members by calling for members to
disagree and confi'ont one another (Priem & Price, 1993; Zander, 1994).

The next section reviews empirical research attempting to improve decision
making through the use of techniques that stimulate controversy. In particular, this
review focuses on two techniques known as Devil’s Advocacy (DA) and Dialectical
Inquiry (DI).

Controversy-Based Interventions in Individugl Decision Mgking

A popular approach intended to improve goup performance in ill-structured
situations has focused on stimulating disagreement among group members (Schwenk,
1990; Taylor, 1992). Although a number of studies have examined the effects of
introducing controversy into the group decision making, most research has focused on the
impact on group performance while ignoring the processes through which controversy has
its efi‘ects. Several studies have measured the number of assumptions and

recommendations identified by groups using controversy-based techniques in comparison

29
to other approaches, no study has attempted to measure the degree to which controversy
stimulates g‘oup members to share information. The upcoming section provides an
overview of research on the two controversy-based decision aids that have been
extensively studied: Devil's Advocacy and Dialectical Inquiry. Empirical studies of DA
and DI can be divided into three general categories: (1) early case studies centering on
the effectiveness of D1 in actual organizational decisions, (2) laboratory research focusing
on DA and DI as methods for creating cognitive conflict within individuals and (3)
laboratory research examining DA and D1 in the context of group decision making.

DA vs. DI. Two particular methods have been extensively studied as mechanisms
for introducing controversy into group decision making - Dialectical Inquiry (DI) and
Devil's Advocacy (DA). Both techniques are predicated on the assumption that top
management will benefit from considering multiple alternatives or options related to
achieving some organizational goal. Although sharing this core assumption, as examined
in the literature, DA and DI differ in terms of how they bring alternative approaches into
consideration. DA essentially requires that the assumptions and data underlying a
proposed plan of action be identified and subjected to criticism. The plan is then revised
on the basis of this criticism and presented again. Iterative cycles of revision and critique
are conducted tmtil all criticisms have been satisfied. On the other hand, DI requires that
a "comterplan" be identified based on assumptions diametrically opposed to the original
plan. The "plan" and the "counterplan" are presented to the group in the form of a
structured debate and a set of "surviving" assumptions is identified and used as the basis
for synthesizing the two plans. As a result, criticism has been seen as the crucial element
defining a DA approach while construction of a second "counterplan" has been viewed as
the defining essence of D1 (Mason, 1969; Cosier, 1978).

Mason (1969) and Mitrofl' and Mason (1981) argued that both DA (critique) and

DI (diametric cormterplan) should improve group decision making over the presentation

30
of one plan by experts from a single functional area. Further, Mason suggested that DI
should yield better group outcomes than DA because it provides a credible alternative to
the existing plan and allows for a synthesis of the best ideas fiom both plans. According
to Mason (1969), the DI structure allows the thesis (first plan) and antithesis (second
plan) to be creatively merged into a constructive synthesis by top management (Mason,
1969; Mason & Mitrofi‘, 1981). On the basis of early research using a case study
approach (see below), Mason identified DI as the technique of choice for use in
improving decision making in organizations.

Although this paper is concerned with groups and the processes underlying
efiecfive group decision making, most of the early research comparing DA and DI
involved individual decision makers instead of groups. In the typical individual-level
study, individual decision makers were exposed to a short "plan," often presented in
writing, and then either exposed to counter-arguments or a second "counterplan." The
intention of such a manipulation was to induce controversy within the decision maker
and so stimulate the consideration of alternatives (and possibly allow for a creative
synthesis). Unfortunately, empirical research using DA and DI with individual decision
makers is severely restricted in terms of its applicability to group decision settings.
However, the majority of research on DA and DI has taken this approach and conclusions
have been made concerning the relative merits of these two techniques with little regard
to the distinction between individual and goup decision making. The result is a goat
deal of confusion in the literature (Schweiger, Sandberg & Rechner, 1989; Schwenk,
1990). Therefore, we review all these studies both as a means of identifying and
potentially clarifying the confusion in the literature.

Early field studies of D1. Early studies of conflict-based decision aids focused on
examining the efl‘ectiveness of D1 in actual organizational decision making. In support of

his recommendation for the use of D1, Mason (1969) presented a case study of decision

3 1
making at RMK Abrasives, a real organization. Mason obtained a strategic planning
document from the company's planning department, identified the assumptions
underlying the plan and then created a second plan based on assumptions counter to those
in the first plan. Company managers reported having favorable attitudes towards DI and
indicated that they felt the plan generated using DI was superior to the one they would
have generated based on the recommendations in the first plan only. In a similar study,
Laurenco and Glidewell (1975) used D1 in the resolution of a conflict between a
television station and its corporate headquarters over the degree of control to be exercised
by corporate headquarters. According to the authors, the issue was resolved with a
constructive compromise that was mutually satisfactory.

Mitrofl‘, Barabba and Kilmann (1977) examined the use of D1 in an actual
organization, the Bureau of Census in Washington, DC. Forty-five employees were
clustered into five homogeneous groups and directed to produce planning reports
suggesting new directions for the Bureau. After the five groups produced their difl‘erent
plans, one representative from each group was included in an executive group that
produced a final integrative report. Mitrofi‘ et al. note that the final report was
characterized by several of its members as being both "exciting" and "innovative."

Emshoff and Finnel (1978) studied a form of D1 known as "strategic assumptions
analysis" at a firm they called "Basic Materials." A planning group at Basic Materials
then utilized strategic assumptions analysis to revise an existing strategic plan, and
according to Emshofi‘ and F innel, the resulting plan included a more thorough analysis of
the data and a revised strategy superior to the old one.

Finally, Mitrofi‘, Emshoff and Kilmann (1979) used a modified version of D1
(strategic assumptions analysis) with three mum of managers attempting to decide on a
pricing decision in a drug company. Strategic assumptions analysis was utilized to

examine competing assumptions, identify an expanded set of alternatives, and arrive at a

32
pricing policy that was characterized by the authors as better than that which would have
been adopted without the use of strategic assumptions analysis (DI).

Although these studies provided some support for the claim that structured
conflict in the form of D1 should be used in actual organizations, a number of reviewers
noted the need to delay the conclusion until DI could be examined under more controlled
conditions and contrasted it with alternative approaches to decision making (Cosier,
1978; Schwenk, 1980; Schwenk, 1982). Shortly thereafter, research on conflict-based
interventions turned to the laboratory utilizing a technique known as "Multiple Cue
Probability Learning" (MCPL).

Mom resegrch using MCPL tasks. Laboratory research on structured
conflict has been greatly influenced by a watershed study conducted by Cosier (1978).
Cosier established a paradign for studying the relative effectiveness of DA and DI using
individual decision makers and a relatively simple multiple-cue probability (MCPL)
learning task. Over the next five years, numerous studies of DA and DI were conducted
with individual decision makers and Cosier's (1978) MCPL task. Conclusions drawn on
the basis of these studies have been very influential (e.g., Schwenk, 1990). These studies
are reviewed here because of their prominence in the literature and corresponding
contribution to the confiision surrormding the merits of DA and D1 in group-level
settings.

In that first study, Cosier (1978) had individuals predict a price-to-eamings ratio
(PIE) for a hypothetical firm using three cues and operating in three different decision
contexts ("world states"). Three statistically-independent cues were used to predict the
P/E ratio: the firm's current ratio (X1), inventory turnover (X2), and debt-to-equity ratio
(X3). Prediction was studied in the context of three different world states with different
profiles of cue-criterion relatiOnship. In State 1, the cues were correlated with Y in the

following manner: rxly = .80, rxzy = .50 and my = .20. In State 2, all three cues correlated

33
.50 with the criterion. State 3 was the opposite of State 1, with rxly = .20, rxzy = .50 and
r,‘3y = .80. Each participant made 20 predictions for each of three different profit centers
of a larger organization, receiving feedback after each decision. Each profit center served
as a different "world state."

Cosier (1978) operationalized the single-plan "expert" (E) approach by
giving participants the written recommendations of an imaginary "expert" who advocated
paying most attention to X1, moderate attention to X2 moderately and only slight attention
to X3. The DA criticism stated that there was reason to believe the first expert's
assumptions were dubious and the recommended weighting inapprOpriate. In the DI
condition, the second expert also noted that there was reason to believe the first expert's
recommended weightings were not correct and suggested a weighting scheme
diametrically opposite that of the first expert (i.e., pay most attention to X3, moderate to
X2 and little to X1).

Cosier (1978) examined the between-subjects effect of inquiry method (DA, DI,
and E), and the within-subj ect effect of decision context (world states 1-3) on the
accuracy of predicting the P/E ratio. He found no effect for inquiry method or decision
context but did find that the two interacted. In State 1, E participants had significantly
less judgnental error than either DA or DI participants, but this situation was reversed in
State 3, where DA participants were significantly better at predicting than either the E or
DI participants. No difi'erences were observed between methods in State 2.

There is nothing particularly remarkable about the interpretation of this
interaction. In essence, when the first expert's recommendations were right and
participants only received this expert's view, they tended to do better. When the true state
of the world was different from the recommendations of the first expert, participants did
better when they were told the first expert was wrong (DA) or given the correct set of
weights by the second expert (D1). The superiority of DA in yielding higher quality

34
decisions in State 3 (when the first expert was wrong) was viewed as providing support
for the superiority of DA over DI.

A number of other studies (e.g., Cosier, 1980; Schwenk & Cosier, 1980;
Schwenk, 1982; Schwenk, 1984a; Schweiger & Finger, 1984) have used the MCPL
approach to study the E, DA and DI operationalizations developed by Cosier (1978).
Cosier (1980) included the effects of two individual difl‘erence variables, self-determined
goal difficulty and goal relevance. Schwenk and Cosier (1980) added an additional DA
condition in which the critique was framed as "emotional" and "carping." Schwenk
(1982) looked at a combined DADI condition in which DI participants received a critique
of the first expert's plan in addition to the counterplan and also examined the role of
ambiguity tolerance as a potential moderator variables of the inquiry method-performance
relationship. Finally, Schwenk (1984a) examined still another variation on the DI
treatment, DI+, where participants received a short explanatory statement intended to
reduce confusion over the receipt of conflicting plans. He also examined another
potential individual differences moderator, task involvement.

The inquiry method by decision context interaction found by Cosier (1978) was
found in each of these later studies in similar form, except later studies did not
consistently find E participants to be superior in State 1 (when the first expert was
correct) and sometimes formd that DI resulted in better outcomes in State 3 (when the
first expert was wrong). No study found a main effect for inquiry method or decision
context. Cosier (1980) found DI participants to be significantly more accurate in State 3
(cormterplan correct) compared to their performance in States 1 and 2. Schwenk and
Cosier (1980) fomd E participants performed significantly better than both DI and DA
individuals in State 1 and while both DA and DI participants outperformed E in State 3.
Schwenk (1982) found that, in State 3, all structured conflict methods (simple DA, simple
DI and DADI) resulted in significantly better judgmental accuracy than E. Also, he found

35
that high ambiguity-tolerance individuals in the simple DA and DADI conditions
performed significantly better than their high ambiguity—tolerance counterparts who
received simple D1 or E. Similarly, Schwenk (1984a) found that all of the various
conflict-based techniques (DA, DI and DI+) outperformed E in State 3 but did not difi'er
among themselves. In addition, highly-involved DI and DI+ participants exhibited
significantly better judgmental accuracy than highly-involved E and DA participants
across all states.

Sgdies of DA/DI gs_ipg non-MCPL taikg In addition to studies using the MCPL
paradigm to assess the relative merits of DA, DI and E, several studies have attempted to
assess the relative eflecfiveness of DA and DI on individual decisions using tasks more
complex than the relatively well-structured Cosier (1978) MCPL task. In some cases,
these studies have also used "real-world" decision makers instead of students, further
increasing the strength of their generalizability.

Cosier, Ruble and Aplin (1978) attempted to assess the relative efficacy of DA
and DI using an eight-period business simulation configured to operate under three
difl‘erent world states similar to the MCPL decision contexts. Four inquiry conditions
were examined: DA, DI, E and a control condition. In State 1, the DA treatment led to
significantly better prediction performance over the DI and control treatments while in
State 2, the control treatment resulted in the most effective performance, yielding
significantly better prediction than the DA, DI and E conditions.

Cosier and Aplin (1980) gave 32 United Way planners a case study involving an
actual United Way agency and asked them to prepare a set of recommendations.
Participants in the control condition got no added information, while participants in the E
condition received a plan generated by the authors of the study. In addition to this plan,
DA participants received a critique of the plan and DI participants received a counterplan

based on difl'ering assumptions. Plans were evaluated by three judges along six

36
dimensions: internal consistency, consistency with the environment, available resources,
satisfactory risk, time horizon and workability. An overall evaluation was also given.
DA was judged superior to D1 on only one criterion, satisfactory risk.

Schwenk and Thomas (1983) used a business scenario in conjunction with actual
managers in examining the relative effectiveness of three inquiry methods (DI, E and a
control condition). The task was the Sweetsa case, requiring participants to specify
actions with regard to harvesting operations in a plantation system. B was
operationalized by giving participants an analysis of the case and a corresponding set of
recommendations generated by an "expert consultan " that would have resulted in a gain
of $26,000 had it been routinely implemented. DI was operationalized by providing
participants with the analysis and recommendations of the first expert, plus an additional
analysis and set of recommendations provided by a second expert whose
recommendations would have resulted in a gain of $4,000 more than the first expert's
plan. Those in the control condition received no additional information.

Results indicated that participants in the DI condition generated significantly
better solutions than those participants in the E condition, with the average DI solution
yielding roughly $32,000 more than the average E solution. In addition, four of the nine
DI individuals identified the optimal solution, while none of the eight E participants were
able to do so. The majority of participants in the control condition recommended
solutions that were deemed "infeasible" due to one or more infractions of the task "rules."

Schwenk (1984b) also examined the relative efl'ects of DA, DI, E and a control
condition with individual undergraduate business students acting as individuals and a
business scenario involving the operations of a fictitious soft drink company. Information
provided in the case centered around feasibility of two potential strategies (acquisition of
winery, development of a new soft drink). Control participants received no added

information other than that contained in the case study while B participants received a

37
"planning committee report" recommending acquisition of the winery. DA participants
received the planning committee report plus the critique of a second planning committee
which questioned the analysis and recommendations of the first committee report.
Finally, the DI participants received the first planning committee report as well as a
second planning committee report offering a "difierent recommendation" (p.266). In
addition to these four inquiry methods, Schwenk examined the effect of presentation
format for the inquiry method instructions (i.e., in writing or by videotape).

Results indicated an inquiry method by presentation medium interaction, with DA
participants generating significantly more strategic alternatives than B or DI participants
when instructions were presented in writing, while control participants generated
significantly more fimctional area alternatives than either of the two conflict-based
approaches. No difi‘erences were observed across inquiry methods in the videotape
condition. Schwenk also reported a significant chi-square showing that participants' final
recommendations depended on the inquiry method they received. However, since the
task provided no qualitative means of distinguishing the various recommendation
alternatives, this is difiicult to interpret. Finally, Schwenk reported that participants in
both the DA and DI conditions indicated greater satisfaction than did participants in the E
condition.

Cosier and Rechner (1985) extended previous research on DA and DI by using a
complex business simulation (SIMQ) and samples of both undergraduates and real-world
managers. They also examined the efl‘ects of experience with inquiry methods over time
by having participants make decisions for four decision periods with the same inquiry
method. Fom' inquiry conditions (DA, DI, E and a control) were examined in conjunction
with several method variables. As in previous research, control subjects received no
additional information while B participants received a comprehensive planning report

plus recommendations for plant operation. DA participants received a planning report

38
plus a critique of the report that essentially questioned every recommendation made in the
initial report, while DI participants received the first planning report plus a second
planning report recommending a course of action diametrically opposed to the
recommendations of the first report. Two planning reports were generated, with half of
the subjects getting Report 1 (R1) as their "expert" report and half receiving Report 2
(R2). For DI participants, one of the plans was received fiom the first expert and the
other from the second expert. The effects of inquiry method were examined over four
operating periods. The second study utilized exactly the same design except with a
sample of actual managers instead of students.

Results of the first study indicated an interaction between inquiry method and
planning report received from the first expert (R1 v. R2). Analysis of the interaction
indicated that DA participants earned significantly more revenue in the simulation than
the DI participants when R2 was presented first. Thus, in the first study with
undergraduates, the benefits of DA were contingent on the characteristics of the first plan
presented. A supplementary data analysis indicated that DA may have tempered the
willingness of participants to attempt the more difficult and complex recommendations of
R2 to a greater extent than DI. On the other hand, the second study utilizing 30 actual
managers found no difierences across inquiry condition. The only significant finding was
that, unlike the students, managers improved their performance over the first three
operating periods and, in general, earned more revenue than the students.

8mm of individual-level research. The studies reviewed above all focused on
stimulating conflict within an individual using either MCPL or more ill-structured tasks.
In general, findings have been mixed and, as a result, conclusions about the relative
superiority of the methods have been heavily qualified.

Studies using the Cosier (1978) MCPL paradigm to examine the effectiveness of
DA and DI inquiry methods tended to replicate the earlier study with minor extensions,

39
usually adding an additional inquiry method condition or including some individual
difference variable hypothesized to mediate or moderate the inquiry method-performance
relationship. For the most part, studies have found that in State 3 where the first expert is
wrong and the critic is correct, DA and DI tend to result in fewer judgmental errors than B
(where errors are defined as the deviation of judgments from "actual" values). In a
corresponding fashion, in State 1 where the expert is right and the critic is wrong, E
participants tended to do better.

Schweiger and Finger (1984) reviewed the results of seven studies comparing DI
and DA in controlled laboratory settings and concluded that no method of inquiry had
been shown to be more efl‘ective than any other. Across all studies, the vast majority of
mean comparisons were nonsignificant and those that were significant did not show a
consistent pattern. They concluded that there was no firm support for the relative
superiority of either DA or DI. Individual-level studies using more ill-structured tasks
have produced similarly mixed results, with DA superior to DI in some studies, DI
superior to DA in others, and neither superior to each other in most cases.

Criticisms of individuA-level research. Laboratory studies intent on examining
the efl‘ectiveness of structured conflict methods in conjunction with individual decision
makers has been subjected to three major criticisms: (1) Tasks used have not been
suficiently complex to generalize to real-world settings, (2) Operationalizations of
structured conflict (i.e., DA and DI) have not generated conflict within individuals and
(3) Structured conflict methods are not intended for use with individuals and should be
studied in group-level settings. The first criticism has been readily acknowledged (i.e.,
Schwenk & Thomas, 1983; Cosier & Rechner, 1985), but the latter two objections have
stirred some measure of debate in the literature.

With regard to capturhg the essence of conflict with past operationalizations,
Schweiger and Finger (1984) suggested that, instead of creating a real sense of "conflic ,"

40

past MCPL operationalizations of DA and DI probably just created confusion in the
minds of participants. They suggested that, given written recommendations fiom
imaginary "experts" who neither provided reasons for their recommendations nor were
present to address questions, participants might have simply attended to one expert and
discounted the input of the other. They hypothesized that the lack of consistency in
earlier findings using the MCPL task may have been due to primacy or recency effects
resulting from participant confusion.

As a means of testing this, Schweiger and Finger (1984) replicated the classic
Cosier (1978) study while creating two conditions for DA and DI wherein the content and
order of expert/critic presentation was reversed. Consistent with their predictions, an
order effect was found suggesting that participants attended to one expert (the first) while
ignoring the advice of the other. .

However, the most damaging criticism leveled against MCPL studies of DA and
D1 is the claim that the entire approach fails to capture the conditions for which DA and
DI were designed: group settings (Mitrofl‘ & Mason, 1981; Mitrofi‘, 1982; Schweiger &
Finger, 1984; Schweiger et al. 1986). Mitrofl' and his colleagues (Mitrofi‘ & Mason,
1981; Mitrofi' & Emshofl‘, 1979; Mitroff, 1982) have been especially vocal in pointing out
that DA and D1 are intended as decision-aids for use in social situations. They claim that
through intense and sometimes heated discussion and debate, managers learn from the
critical re-appraisal of strongly held beliefs and assumptions. According to this position,
the well-structured MCPL task is not an appropriate task for studying group-level
decision making, and comparisons between methods are thus meaningless.

It is important to note that this last criticism is conceptual in nature and cannot be
addressed empirically. The value of much of the published literature on structured
conflict is at stake. Clearly, there is no "ri t" answer to this question, but the growing

consensus is that, for the most part, the criticism is valid (Schweiger et al., 1989;

4 1

Schwenk & Cosier, 1993). Rather than completely discounting the numerous studies
which have adopted the individual level approach, it seems more reasonable to ask the
question of each study, "How confident can one be that the results will generalize to real-
life strategic decision making settings?" At least four factors may affect this: (1) The
adequacy of the construct operationalization in terms of stimulating "conflic " within
decision makers, (2) The structure and complexity of the task and (3) The sample
employed in the study (managers v. students) and (4) The level of analysis (group v.
individual). Table 1 describes the studies of structured conflict that have been done to
date, and provides summary information on level of analysis, task type, sample, and
results obtained.

By their very nature, studies of controversy-inducing techniques such as DA and
DI involving individuals cannot address the efl‘ects of such methods on group process or
group outcomes. In particular, it is not possible to examine the effects of introducing
controversy on intragroup conflict or information sharing. At best, these individual-level
studies suggest that stimulating controversy within individuals may result in the
consideration of difi‘erent information and/or alternatives. The few studies that have
examined DA and D1 in group settings clearly provide the strongest base for making
statements about the relative efl‘ectiveness of conflict-based decision aids to real-life
strategic decision making. To this point, four studies have heeded their call, and we now

turn to these studies.

42

Table 1.

Empirical Studies of Controversy-Based Decision Aids.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Author-Year Level Task Sample Results
Cosier (1978) Ind. MCPL Students DA > DI (S3)
Cosier (1980) 1nd. MCPL Students Dl (S3) > D1 (81 & 82)
Schwenk & Cosier (1980) Ind. MCPL Students DA = DI
Schwenk (1982) Ind. MCPL Students D1 = DA = DADI > E (83)
Schwenk (1984a) Ind. MCPL Students D1 = DA = Dl+ > E (S3)
Schweiger & Finger 1nd. MCPL Students
(1984)
Cosier, Ruble & Aplin Ind. Case Managers DA > D1 (81)
(1978 Study
Cosier & Aplin (1980) Ind. Case Students DA = DI
Study
Schwenk & Thomas Ind. Case Managers DI > Control
(1983) Study
Schwenk (1984b) Ind. Case Students DA > Dl (written instructions
Study only)
Cosier & Rechner (1985) 1nd. Bus. Students DA > Dl (students receiving R2
Game Managers first only)
Chanin & Shapiro (1984) Grp. Bus. Students D1 = DA > C8 (pert)
Game
Schweiger et a1. (1986) Grp. Case MBAs D1 = DA > CS (pert)
Study D1 = DA < CS (affect)
Schweiger et a1. (1989) Grp. Case Managers D1 = DA > CS (pert)
Study D1 = DA < CS (affect)
Schwenk & Cosier (1993) Grp. Case Students DA > C8 (pert)
Study

 

 

 

 

 

 

43

Controversy-Based Interventions in Group Decision Making

Chanin and Shapiro (1984) conducted the first published study of structured
conflict on group decision making using a moderately complex business simulation, the
Executive Game. They used four person groups along with specialized roles for each
group member corresponding to the functional diversity in strategic decision making
settings (one president and three vice presidents). Each of the vice presidents represented
a department within the organization, representing to some degree the diverse
backgrounds of individuals typically involved in strategic decision making. Three inquiry
methods (DA, DI and a control condition) were examined for their effect on several
quantitative performance variables generated by the simulation (i.e., industry ranking,
retmn on investment, net profit).

DA and DI were operationalized by having each vice president individually
prepare an operational plan from the data base of performance information generated by
the simulation. In the DI condition, each plan was then presented to the entire group
along with corresponding assumptions and supporting data. After all three plans had
been presented, groups were instructed to conduct a general discussion of the pros, cons
and underlying assumptions for all three plans. Following this, groups were told to agree
on a final set of assumptions and jointly develop an integrated strategic plan. The manner
in which DA was operationalized is less clear. The authors confined their description of
DA to noting that it involved the following four-step process:

1) development of strategic and operational plans (forecasts);

(2) plan presentation at the management briefing session;

(3) management critique of the plan; and

(4) development of a final plan (pp. 165-166).

Control groups were allowed to operate freely and make decisions using a structure of
their own choice.

Chanin and Shapiro (1984) trichotomized the 51 teams in their study into high,

medium and low categories on two performance variables, industry ranking and return on

44
investment, for each of the three simulation years. For each year and index, there were
more DI groups in the high performance category than either DA or control groups. DA
groups were noted as being most prominent in the medium performance group while
control groups appeared in the low performance category more often than either D1 or DA
groups.

Noting that the previous categorization scheme could not be tested for
significance, t-tests were also conducted to compare the three inquiry methods on a
number of performance variables across the three decision periods in the game. With
respect to the 15 t-tests possible comparing each method to another over the three
decision periods, DI was significantly better than the control condition for nine
comparisons (60%), while DA was better than the control group for only two variables
(19%). DI was superior to DA on two comparisons as well (19%), both in third and final
year of the simulation. The control condition did not produce higher outcomes than D1 or
DA at any time. However, given the likely interdependence of the 15 measures, it is
dificult to interpret these percentages in a straightforward fashion.

0n the basis of these findings, Chanin and Shapiro concluded support for a "very
strong and statistically significant difierence" between DI and the control condition.
Furthermore, they noted that the study permitted classification of DI as high, DA as
moderate, and the control conditions as low performance problem-solving technologies.
These results suggest that DI does produce superior group outcomes compared to DA, but
given the methodological issues noted above, this conclusion must be qualified.

In the second study on structured conflict in a group setting, Schweiger et al.
(1986) compared the two conflict-based methods (DA and DI) against a consensus-
seeking (CS) approach. Using a case analysis scenario (the Leitch Quality Drug
Company), Schweiger et a1. operationalized DA and DI by splitting four-person groups
into two—person subgroups, with one subgroup in both DI and DA assigned the

45
responsibility of developing and presenting recommendations to the other subgroup along
with all supporting assumptions and data.

In the DI condition, after receiving the assumptions of the first subgroup, the
second subgroup then formulated a counterplan based on assumptions that negated those
of the first subgroup. Following this, both subgroups presented their assumptions and
recommendations and supporting data to the other subgroup in a structured debate setting.

In the DA condition, the second subgroup was instructed to prepare a critique of the plan
recommended by the first subgroup. After receiving the critique from the second
subgroup, the first subgroup modified their assumptions and recommendations and
presented the revised plan to the critiquing subgroup. This cycle was repeated until the
plan was approved by the second subgroup. The consensus-seeking (CS) approach was
implemented by giving groups a variety of instructions generally attempting to get
members to be skeptical, noncompetitive, reasonable and flexible.

Schweiger et a1. (1986) measured a number of process and outcome variables,
including the number, validity and importance of identified assumptions as well as the
quality of final recommendations. These variables were measured by having two judges
independently code transcripts generated from audiotape recordings of each group as they
solved the case. They also measured several afl'ective variables such as satisfaction,
desire to work with the group again in the future, critical reevaluation of assumptions and
acceptance of the group's decision.

Schweiger et al. (1986) found that although DA and DI groups did not identify
more assumptions than CS groups, both methods led to the identification of assumptions
characterized as more valid and more important than those of the consensus groups. DA
and DI groups also generated higher quality recommendations than CS groups, and
resulted in more critical reevaluation of assumptions at the individual level. With respect

to comparisons between DI and DA, DI groups were only superior to DA on one process

46

variable (validity of assumptions) and there was no significant difference between the two
methods with regard to the quality of recommendations. On the other hand, groups in the
consensus condition expressed more satisfaction with the task, greater willingness to
work together in the future and greater acceptance of the group decision than did groups
in the DI and DA conditions.

Schweiger et a1. (1989) followed up their earlier study by examining DA, DI and
CS in conjunction with groups of real-life managers and multiple decision situations. In
addition to the Leitch Quality Drug Company case, Schweiger et a1. (1989) added a
second case analysis, the Hudepohl Brewing Company case. Managers were randomly
assigned to four person groups in one of three inquiry method conditions (DA, D1 or CS)
and used the same decision making approach for their second decision as well. DI, DA
and CS were implemented in the same fashion as the Schweiger et al., (1986), and the
same process and outcome variables were measured along with the time taken to
complete each decision task. Each session was again tape recorded.

The results of Schweiger et a1. (1989) were consistent with Schweiger et a1.
(1986) and provided further support for conclusions drawn fi'om the earlier study. As
before, the number of assumptions identified by groups was not related to inquiry
method, but the validity and importance of identified assumptions was higher for both
DA and DI groups compared to the CS groups. Also, both DA and DI groups again
produced superior recommendations and greater individual critical reevaluation of
assumptions than CS groups. Of particular significance, there were no differences
between DA and DI for any measured variables, including validity of assumptions. With
respect to meeting time, CS groups took significantly less time than DA and DI groups in
the first decision task, but not in the second. All groups produced higher quality

assumptions and recommendations and took less time on their second decision task.

47

The pattern of results with respect to the afiective variables was similar to
Schweiger et al. (1986) for the first decision, but not for the second task. In general, CS
groups reported superior afl‘ective responses for the first decision, but not in the second.
There was no efl‘ect of inquiry method on satisfaction but all individuals reported higher
satisfaction with their groups in the second task. In the first decision, CS group members
reported significantly greater acceptance of the group decision than did individuals in the
DA and DI conditions, but this advantage was narrowed and became non-significant in
the second decision.

Schwenk and Cosier (1993) appear to have been persuaded by the arguments of
Schweiger and his colleagues that DA and DI were meant to be employed as aids to group
decision making. In the first study by either author involving groups, Schwenk and
Cosier (1993) compared the relative effectiveness of DA and CS methods while
measuring the degree of agreement within each group with respect to the groups
objectives. The task used was the same case study as that employed by Schwenk (1984)
involving a soft drink company focused around the major decision of whether to acquire a
winery, develop a new soft drink, do both or do neither.

Similar to the Schweiger et al. (1986; 1989) studies, Schwenk and Cosier (1993)
measured several process and outcome variables, including the number and quality of
assumptions identified, the quality of recommendations, the degree of critical self-
evaluation, as well as several afiecfive variables (commitment to decision, desire to work
again in future). Although the analyses indicated some efl'ect of within-group agreement
on objectives, no significant difference due to inquiry method was found for any of the
performance variables (i.e., number of assumptions, quality of assumptions or quality of
recommendations). However, as in Schweiger et al. (1986; 1989), participants in the CS
condition expressed more satiSfaction and desire to work again as a group than did DA

participants.

48

Effects of controversy-based decision aids. A great deal of time and effort in the
last 25 years has gone into talking about, arguing over and (to some extent) studying the
degree to which controversy-based decision aids are useful tools for increasing the
likelihood that organizations will make good strategic decisions (Schwenk, 1990).

For the most part, the most of the literature on DA and DI has focused on
answering the question of the relative eflicacy of the two techniques. In reviewing the
literature, Schweiger et al. (1989) noted that the answer to this question has been clouded
by variation in tasks, experimental samples and, most importantly, the level of analysis
used across studies. Early studies conducted in the field using actual managers in real
organizations suggested that DI was a useful tool for exposing top management to
alternative approaches (Mitrofl‘ & Mason, 1981; Mitrofl‘, 1982). Laboratory research on
DA/DI using individuals resulted in mixed findings. An early review of laboratory
comparisons of DA and DI using the MCPL task noted no clear superiority for either DA
or DI (Schweiger & Finger, 1984). Finally, three recent studies conducted with groups
have all formd that neither DA nor D1 is better than the other with respect to group
performance or individual afl‘ective outcomes.

Schwenk (1990) conducted a meta-analysis of the efi‘ects of DA, DI and E on
group performance using 16 published studies and 17 effect sizes. On the whole, DA was
found to be superior to the E approach while DI was not. Further, DA was not formd to
be reliably more efl‘ective than DI. According to Schwenk, the meta-analysis "supports
one clear conclusion which has been disputed in past literature reviews: The DA
improves decision making over an expert-based approac " (pp. 170-171). However, DA
was not found to be better than DI.

Given the growing consensus that strategic decision making usually occurs in
group settings, it is questionable whether the Schwenk (1990) meta-analysis is capable of
addressing the issue of relative eficacy between DA and DI. Only three of 16 studies in

49
the Schwenk (1990) meta-analysis employed groups as their decision making unit. This
calls into question the relevance of the findings to group-level decision making.

Looking at both individual and group-level studies, Schwenk (1990) addressed
this issue by identifying three potential moderators of the efl‘ects of DA, DI and E on
decision making performance - type of subject (students v. managers), decision making
level (individuals v. groups), and type of task (MCPL v. non-MCPL). A sub-group
analysis was performed for the task type moderator, but analysis of the other moderators
was not conducted due to the small number of studies utilizing practicing managers and
groups.

In spite of the few studies examining DA and D1 in the context of group decision
making, these studies enjoy two strong advantages over the remainder of the literature on
controversy-based interventions: (1) They are the only studies conducted at the
appropriate level of analysis and (2) Their findings are very consistent. Three different
studies of DA and DI using groups have found neither method to be superior to the other
(Chanin & Shapiro, 1984; Schweiger et al., 1986; Schweiger et al., 1989). At the same
time, all three studies fotmd both DA and D1 to be superior to instructions to simply reach
consensus. A fourth study by Schwenk and Cosier (1993) which only examined DA and
CS also found DA to be better than CS. Further, the two controversy-based methods have
consistently produced lower member satisfaction, acceptance of the decision and desire to
work with the group again in the future than the CS approach, with neither DA nor DI
superior to the other.

Given the striking consistency of the group-level findings, it appears that it is now
time to conclude that DA and DI as traditionally implemented work equally well. The
studies available suggest that it does not make much difference whether DA is used or DI
is used - either method is likely to yield higher group performance outcomes but lower

group afi‘ective outcomes than a corresponding emphasis on seeking consensus. At the

50

same time, we know very little about why these techniques work. Little attention has
been given to the processes through which DA and DI have their effects. Although in
some sense it is enough to know simply that they do improve group decision making
outcomes, both DA and DI introduce complex dynamics to group member interaction that
need to be isolated and understood. It may be the case that not all aspects of controversy
promote effective group decision making. In particular, the evidence we have reviewed
thus far suggests that controversy impacts decision quality both positively and negatively.

Sammy: The Dilemma Surrounding ControversL The relationships between
information sharing, intragroup conflict, and group decision making quality imply a
dilemma for those considering the use of controversy-based intervention techniques in
decision making groups. In group contexts with diverse member composition and
knowledge, it is necessary for individual experts to share relevant information which they
alone have access to in order for there to be any chance of an optimal group decision. At
the same time, interventions centered around stimulating controversy among group
members may very well lead to intragroup conflict within the group. This intragroup
conflict in turn may reduce the ability to integrate all the information shared by individual
members as the group goes about deciding on a collective decision or plan. The few
studies which have examined controversy-based techniques in group decision making
settings have found them to yield more effective decisions than instructions to reach
consensus; however, this may not always be the case. Ifthe level of conflict becomes
high enough, the use of controversy-based techniques may actually lead to lower quality
decision outcomes.

Given the many factors contributing to low information sharing in groups,
prescriptive interventions in group-level decision making must provide some mechanism
which allows group members to overcome the potential hazards posed by groupthink and

biased information sampling. The ideal intervention in such circumstances is one that

5 1
maximizes information sharing among individuals and minimizes the amount of
intragroup conflict generated in the process. Furthermore, this intervention may already
exist.
The Forgotten Role of the Synthesis

In an influential book in the early 19805, Mason and Mitroff (1981) identified
three firndamental components of the dialectical inquiry: (1) a thesis, or favored plan, (2)
an antithesis, or counterplan, constructed from values and assumptions contradictory to
those underlying the thesis and (3) a synthesis, or new world view that incorporates the
best features of both plan and counterplan yet somehow manages to reflect a “worldview”
different from that which served as the basis for either plan or counterplan.

The literature has largely ignored the role of synthesis in dialectical inquiry. In
two studies, Schweiger and his colleagues (Schweiger et al., 1986; Schweiger et al.,
1989) implemented DA and DI by splitting decision making groups into two two-person
subgroups. Schwenk and Cosier (1993) implemented DA in the same fashion. Chanin
and Shapiro (1984) implemented DI in four-person groups by having three members
(departmental vice presidents) prepare, present and debate individual plans while one
member of the group merely participated in the ensuing discussion. To our knowledge,
no study has explicitly assigned the synthesis role to an independent bloc of group
members in the fashion intended by the original descriptions of the dialectical process.

Why is synthesis crucial to the efl'ectiveness of the dialectical inquiry process?
Mason and Mitrofl‘ (1981) suggest that the synthesis embodies a new and “higher”
understanding of the problems or issues faced by the group. The synthesis provides a
deeper and insightful way of thinking about the matter. They explicitly highlight the
transformational character of the synthesis; it incorporates a new worldview difl‘erent
from those that inspired the plan and counterplan. They suggest that the importance of
the synthesis lies in its ability to provide a new perspective that, while borrowing from

52
and similar to both thesis and antithesis, is at the same time fundamentally novel and
distinct.

It is somewhat surprising then to note that empirical studies of D1 have largely
ignored the role of synthesis in the dialectical process. As typically implemented, DI
instructions assign halfof the group to the formation of a plan, the other half of the group
to the formation of the counterplan, and no one to the role of overseeing a synthesis. In
fact, no study was found that explicitly assigned the synthesis role to one or more group
members. In most cases, group members were collectively asked to develop a synthesis
at the conclusion of the structured debate over plan and counterplan. To the extent that
intragroup conflict results fi'om the controversy engendered during the debate, it seems
overly optimistic to expect group members who have been actively arguing for the thesis
or antithesis to put aside their feelings, opinions and views and creatively forge a new
“worldview.”

One of the goals of this study is to provide an examination of dialectical inquiry
the way it was intended to be implemented: with thesis, antithesis and synthesis roles
explicitly assigned to independent sub-groups. At the same time, neither Mason (1969)
nor Mason and Mitroff (1981) has clearly specified the individual behaviors and
underlying group-level processes that promote the formation of a new “worldview.”

In the context of the complex and ill-structured decisions characteristic of
strategic decision making, the synthesis role would seem to involve a set of behaviors that
promote efl'ective group decision making. A well-known distinction in the literature on
groups and teams involves task and maintenance functions (Levine & Moreland, 1991;
Salas et al., 1992). Behaviors with a task orientation are concerned with maximizing
production or meeting the group’s goals; behaviors with a maintenance orientation are
most concerned with promoting/maintaining positive interpersonal relations and harmony
within the group (Benne & Sheats, 1948; Thiabut & Kelley, 1959). The task and

53
maintenance functions now appear to have been accepted as fundamental dimensions for
describing the behavior of group members. The definition of the synthesis role as defined
by Mason and Mitroff (l 981) seems closely associated with the "task" function but, when
implemented in groups, may also serve a maintenance function as well.

Summarizing the results of earlier research, Patton et a1. (1991) identified a set of
individual behaviors corresponding to these two basic dimensions. Behaviors that serve a
task orientation include such things as initiating structure, stimulating communication,
clarifying communication, summarizing, and consensus-testing. “Initiating structure”
behaviors include proposing objectives to the group, introducing procedures, developing
an agenda, and suggesting the group move on to new topics. “Stimulating
communication” refers to direct requests for other group members to provide information
or opinions. “Clarifying communication” pertains to efl‘orts to reduce confusion by
asking questions or interpreting ideas. “Summarizing” behaviors provide a review of
what has been said or accomplished so far. “Consensus-testing” behaviors are intended
to provide some indication of the extent to which group members agree on what has been
said or proposed.

In domains other than psychology, researchers have identified similar sets of
behaviors intended for use by facilitators or mediators intended to help groups under
stressful decision making conditions. Martin (1983) listed five important behaviors for
such circumstances: (1) Reflecting, (2) Silence and Attentive listening, (3) Asking for
specifics, (4) Making I-statements, and (5) Focusing on areas of agreement. Frost and
Wilmot (1978), in discussing tactics for intervening in group processes, noted the positive
impact of behaviors such as (1) Being descriptive instead of judgmental, (2) Encouraging
specificity, (3) Providing feedback, (4) Setting and keeping to an agenda and/or
associated time limits, (5) Comparing and restating positions, (6) Summarizing, (7)

Providing information, (8) Calling on persons in a non-threatening way and (9) Forging

54
commitment to negotiated plans. A common theme uniting these behaviors is their focus
on drawing out information and structuring group activities while without provoking
interpersonal conflict.

Although possible to categorize these behaviors at various levels of specificity, it
is possible to identify four general behavioral dimensions that serve to promote the
eflecfive processing of information in groups fi'om these lists. These dimensions are:

(1) Reflecting/summarizing what has been said by other group members, (2) Asking other
group members clarifying questions or for specific information, (3) Integrating ideas and
recommendations, and (4) Focusing/ structuring group activities and discussion items.
These four behavioral dimensions comprise the basic elements of a broad construct which
can be called process facilitation. Process facilitation is defined as a set of behaviors
employed by one or more individuals within a group intended to help a group structure its
activities and use its informational resources as effectively as possible.

It seems likely that these behaviors will have positive effects on both information
sharing and intragroup conflict in groups involved in the dialectical inquiry process.

With regard to information sharing, process facilitation may promote information sharing
in at least three difl’erent ways: (1) As post-hoe responses to summarization attempts,

(2) In response to direct questions asking for clarification of confusing points and (3) As
support for integrative directions proposed by the synthesizing individual(s). Process
facilitation should also impact the level of intragroup conflict that develops within
strategic decision making groups by de-emphasizing the potentially confrontational
situations that develop around the acceptance and support of member ideas. By providing
a neutral “third party” influence, the synthesis role in D1 can dilute the competitive focus
instilled by the debate process, and provide a rubric for compromise, position change, and

backing down without loss of face.

55

Integration

This paper began by defining strategic decisions as complex, important, ill-
structured problems or issues facing organizations. Although organizations confront
many complex and important decisions, strategic decisions are distinctively characterized
by their relevance to the entire organization, multifunctional representation, and lack of
structure (Taylor, 1992). These characteristics imply that strategic decisions will involve
large amounts of relevant data and multiple perspectives with regard to how objectives
can best be attained (Tjosvold, 1985). As a result of this low structure, high information
load and cross-functional relevance, strategic decisions are often made by groups of
individuals from multiple departments within the organization (Koopman & Pool, 1990;
Larson & Christensen, 1993). Koopman and Pool (1990) note that, by gathering
specialists with diverse expertise fi'om diflermt departments, organizations seek to
maximize their chances of high-quality decisions. In other words, for most organizations,
strategic decision making is a group phenomenon.

A Process Model of Group Decision Making

Unfortunately, despite intuitive reasons to expect otherwise, it is clear that
strategic decision making groups can, and do, make poor decisions— sometimes with
disastrous results (Janis, 1972; Janis & Mann, 1977). Like all groups, strategic decision
making groups are susceptible to "process loss" (Steiner, 1972; Hill, 1982). This paper
identified two forms of process loss that may occur in decision making groups: restricted
information sharing and poor integration of shared information.

The focus of attention on information in complex, ill-structured situations is based
on the assumption that a high level of information sharing is a necessary - although not
suflicient - condition for high quality group decisions. Under normal circumstances,

more information is usually better in that it provides more resources from which to form

56
the group product. In these situations where no group member could possibly know
everything needed to create an optimal plan, it is necessary for groups to combine and
integrate information available to multiple members. The failure to consider important
information known to one or more individuals is likely to correspond to a failure to
consider important information. At the same time, in many situations, high levels of
information sharing may not be crucial for good decisions but rather optimal decisions, as
individuals can couch information within high-quality recommendations without
explicitly sharing information. Still, it seems reasonable to believe that, in most
situations, high levels of information sharing will result in maximal decision quality.

Research on "groupthink" (Janis, 1972; Janis, 1982) and biased information
sampling (Stasser & Titus, 1985; Stasser et al., 1989) has called attention to the tendency
for groups not to discuss information known to one (or only a few) group member(s). In
the case of "groupthink," although the empirical evidence is scant, high group cohesion,
external threat and a desire to reach consensus quickly are hypothesized to result in
restricted consideration of information known to the various group members. In the case
of biased information sampling, research suggests that groups tend to spend their time
discussing information known to all group members. Both of these approaches suggest
that low information sharing among group members leads to sub-optimal decision
outcomes.

Although it is clear that high levels of information sharing are necessary for
optimum group decision making, decision quality is likely to be afi‘ected not only by the
amount of information available to the group as a whole, but also the manner in which
that information is used once it has been shared. A second form of process loss can occur
when groups do not adequately utilize information at their disposal. Furthermore, a
review of the literature suggests that intragroup conflict may inhibit the ability of groups
to combine the information at their disposal (T josvold, 1985; Pruitt & Rubin, 1986;

57
Zander, 1994). Although the intrapersonal and interpersonal negative effects of
intragroup conflict have received a great deal of attention in the literature, the effects of
intragroup conflict on group—level information processing have received much less
attention (Tjosvold, 1985; Zander, 1994).

Conflict among group members has been found to produce a variety of afi‘ective,
behavioral and cognitive consequences at the individual level, including anger,
annoyance, frustration, hostility, insults, issue distortion, cognitive simplification and
irrationality. In groups, conflict breaks down relationships, hinders communication,
obstructs problem solving (Tillett, 1991). As group members become agitated and
annoyed during group discussion, they tend to become defensive, close-minded and less
willing to accept the views of others. Emotions become aroused, initial positions tend to
harden and group members become hypersensitive to the threat of losing “face.” The
desire to avoid losing face leads members to forego admitting that someone else might
have a better way of doing things or even a useful recommendation. As certain members
begin to dominate, the information shared by dissenting members of the group may tend
to be left out of the collective plan. Thus, when levels of intragroup conflict are high,
constructive synthesis of differing viewpoints seems unlikely as the information
processing resources available to groups in the form of time and individual attention are
diverted or reduced. Therefore, I propose that:

H1: The level of intragroup conflict in strategic decision making groups will
moderate the effect of information sharing on group performance. When
intragroup conflict is low, a high level of information sharing will result in higher
group performance. However, when intragroup conflict is high, high levels of
information sharing will not result in higher group performance.

The discussion of group processes so far suggests that information sharing among
group members is essential for optimum decision making. Furthermore, we have

hypothesized that intragroup conflict will moderate the relationship between information

58
sharing and decision quality. Figure 1 presents a model that summarizes the discussion of
the processes involved in group decision making for ill-structured task conditions.

In light of the many well-known fiascoes resulting from the failure of decision
making groups to examine questionable assumptions and use all available information, a
number of researchers have advocated the use of techniques that introduce controversy
into decision making groups. Earlier in this paper, conflict and controversy were
distinguished on the basis of their nature (i.e., affective versus cognitive) as well as their
focus (i.e., directed at other members, directed at ideas). Accordingly, the introduction of
controversy is designed to promote the critical examination of assumptions,
recommendations and supporting information as well as inspire search for creative
alternatives. Controversy can lead individuals to become motivated to explore and
understand opposing views and arguments, appreciate the shortcomings of their own
perspective, integrate useful aspects of others’ positions, develop a fresh vieWpoint and,
as a result, make high-quality decisions (Tjosvold, 1985). In strategic decision making
groups, controversy seems likely to promote information sharing, as members are
confionted and called upon to give credible explanations for their positions.

However, at the same time, in strategic decision making environments involving
members with strong, entrenched beliefs and an interest in preserving "face," controversy
seems likely to have a number of unintended negative side-effects as well. For example,
group members may express their opinions directly but close-mindedly or refuse to
acknowledge the appropriateness of another group member's views to avoid the
embarrassment of backing down from a position. As alternative positions harden,
proponents may attempt to find weaknesses in opposing arguments, counterattack and
undercut opposing positions in an effort to impose their own views (Zander, 1994). If
handled inappropriately, disagreement over ideas may soon transform itself into

interpersonal hostility. As a result, I propose:

59

8550 usouwmbfi mo Spam wcufioeoE 2E. ._ 223m

 

 

 

mo:mE.5%5L
@380 A +

 

 

 

 

wccmzm

:osmEeEE

 

 

 

SEED msouwfifiz

 

 

 

 

60

H2: The level of controversy in strategic decision making groups will be positively
related to the level of information sharing.

H3: The level of controversy in strategic decision making groups will be positively
related to the level of intragroup conflict.

The previous two hypotheses form the basis of a dilemma when it comes to
considering the use of controversy-inducing methods such as dialectical inquiry:
although controversy should serve to increase information sharing among group
members, it should also result in high levels of conflict that prevent that information from
being used properly. This existence of this dilemma is indirectly supported by the results
of a growing body of studies that have measured affective outcomes and decision quality
in groups using dialectical inquiry, devil’s advocacy and consensus-seeking inquiry
methods. These studies have tended to find that controversy-based methods such as DA
and DI do produce higher-quality outcomes, but also tend to lead to higher levels of
individual dissatisfaction.

At the same time, studies from a number of domains relevant to group
performance suggest that group members may engage in a number of behaviors that
facilitate interpersonal interaction and the use of information. Process facilitation has
been defined as a set of behaviors that help provide the group with structure in a
procedurally ambiguous environment as well as utilize the informational resources at its
disposal by reviewing where the group has been and identifying where it needs to go. As
a by-product, process facilitation may help quell conflict originating out of the
confiontational atmosphere of the dialectical inquiry process. Whereas the assignment of
all group members to one of two salient "sides" in a debate may accentuate the
confi'ontational aspects of the task, the use of facilitative behaviors should weaken the "us
versus them," "win-or-lose" mindset that might otherwise develop within the group as a
result of the debate. In particular, facilitative behaviors may serve as a springboard for

constructive discussion in the first awkward moments after debate has ended. Group

61

members performing a facilitative function can identify variations of plan or counterplan,
ofi‘er integrative solutions combining elements of both plan and counterplan, probe
advocates for underlying reasoning and data, and generally provide a means for
movement around impasses without the loss of face occurring on either side. By taking
charge of the post-debate process and promoting understanding of the positions that have
been taken, we may expect that an effective synthesis process may reduce the longevity
and severity of conflicts that do arise among members and promote information sharing
through clarification, integration and imposed structure. Thus, I propose the following
hypotheses:
H4: The level of process facilitation in strategic decision making groups will be
negatively related to the level of intragroup conflict.
H5: The level of process facilitation in strategic decision making groups will be
positively related to the level of information sharing.

Figure 2 displays an expanded process model of group decision making showing
the efl‘ects of Controversy and Process Facilitation.
A Prescriptive Model of Group Decision ME

The process model of group decision making just described has the potential to
inform our understanding of the ideal intervention into group decision making. The
literature on group decision making interventions has focused on increasing the
information sharing among group members through the introduction of structured conflict
(i.e., DA and DI). As a result, the ideal group intervention is one that maximizes
information sharing among group members while minimizing the intragroup conflict that
results fiom this sharing.

Given that DI has never been implemented with explicit assignment of the
synthesis role, it is useful to compare the two DI variations (with synthesis role, without

synthesis role) to what may perhaps be the default group decision making norm in

62

mafia—2 :22on 955 me .032 388m < .N ocswfi

 

 

 

00:5.:._O,.—._®L

 

 

 

wccmzm

A
95.5 > 5:328 E

 

 

 

 

 

 

 

35:00 escawmbfi A

1T

 

3.55.5200

 

     

 

 

 

 

 

:OSSEQE .
mmoooi

 

 

 

63
organizations: group member consensus. Research has found that DI implemented
without explicit assignment of the synthesis role ("Traditional DI") has tended to produce
better group decision outcomes but lower affective outcomes for members, possibly as a
result of conflict engendered in the dialectical process. It seems reasonable to expect that
D1 in either form will continue to lead to better collective decisions but also more
intragroup conflict. On the other hand, when comparing the two DI techniques, we might
expect that DI with formal assignment of the synthesis role (“Synthesis” DI) will result in
both increased information sharing and, to some degree, less intragroup conflict. Many
process facilitation behaviors fall very naturally to the member (or members) assigned the
synthesis role in the DI process. The presence of one or more neutral facilitative group
members in the DI process should allow groups to enjoy high information sharing AND
relatively low levels of intragroup conflict. Thus I propose:
H6: Groups employing Synthesis D1 will exhibit higher levels of information
sharing than groups employing Traditional DI, while groups employing Traditional
D1 will in turn exhibit higher levels of information sharing than groups employing
Consensus.
H7: Groups employing Synthesis D1 will exhibit higher levels of intragroup conflict
than groups employing Consensus, but lower levels of intragroup conflict than
groups employing Traditional DI.

Finally, as a result of sharing more information than Consensus-Seeking and
Traditional DI groups, as well as generating less intragroup conflict than Traditional DI
groups, Synthesis DI groups should incorporate the best features of both structured
conflict and consensus-seeking methods. Thus, I propose the following:

H8: Groups employing Synthesis D1 will exhibit higher levels of group performance
than groups employing Traditional DI, while groups employing Traditional DI will

in turn exhibit higher levels of group performance than groups employing
Consensus.

64

Figure 3 displays a model of group decision making that adds the prescriptive
elements discussed above to the expanded process model of group decision making.
However, it is important to note that the model identified here is limited to decision
making groups facing certain task conditions. First and perhaps most importantly, the
model is constrained to situations where task-relevant information is distributed across
members in such a fashion that it is impossible or impractical for one individual to
eflecfively accomplish the task. Indeed, a central feature of the model identified here --
information sharing - makes no sense in situations where all group members possess
essentially the same knowledge of the subject (e.g., juries). Furthermore, given the lack
of feedback loops, this model most directly addresses the process and performance of
groups or teams assembled in an ad-hoc fashion for a particular task. Finally, this model
may be limited to groups or teams that meet in a face-to-face manner. In all likelihood,
the relationships among inquiry method conditions, process variables and performance
might very well be difl‘erent for groups interacting through information-restricted media
(e. g., computer-linked networks). Such media not only limit the richness of interpersonal
interaction but introduce a temporal dimension to interaction not reflected in the present
model.

In reality, as a whole these constraints limit the relevance of the model to ad-hoc
decision making groups interacting in a face-to-face manner and facing broad, diffuse
problems or issues that involve large amounts of specialized information. As such, this
model is most relevant to organizational decision making groups brought together for a
single major “even ” such as creating a strategic plan and/or setting organizational
policies. Therefore, although some group tasks can be accomplished by individuals, this
model applies only to task s that must be performed by decision making groups because

of the physical and information processing limitations associated with individuals.

wet—£2 56609 @280 me .252 033585 < .m oczwmm

 

 

 

 

mocmE._£._oL o\+

wccmsm All
95.5 A »

3.53.530 A]
:o_EE._o%__

 

 

 

 

 

 

 

65

 

 

 

  
  

 

 

 

9.053505—
mmeooi

 

35:00 asewfisc _

 

 

 

 

 

 

 

 

3522 \CEUE \

 

66
Summa_ry

Decision making groups can suffer from at least two types of process loss relevant
to the processing of information known by individual members: failure to share
information with the group and failure to optimally use information which is shared. A
good deal of research on improving strategic decision making has focused on two
particular methods that purport to introduce controversy as a means of promoting
information sharing among group members. Unfortunately, the process of introducing
controversy probably leads to conflict among members, which may then prevent the
optimal integration of information which has been shared.

There is growing evidence in the literature that the most salient distinction
between traditional DA and DI techniques - "critique" versus "counterplan" - may be
relatively tmimportant. On the basis of recent studies using groups, it appears that
critique is interchangeable with counterplan in terms of affecting decision making quality
(Chanin & Shapiro, 1984; Schweiger et al., 1986; Schweiger etal., 1989). At the same
time, these studies have implemented structured controversy in different ways - two two-
person subgroups versus three one-person subgroups and an apparently "neutral" fourth
member. With the benefit of a process model depicting group-level information
processing, I hypothesized that Synthesis DI incorporating facilitative member roles
would result in as much if not more information sharing than Traditional DI and generate
less intragroup conflict.

This study can extend our understanding of the effectiveness of decision making
groups in three ways. First, this study seeks to clarify the role of information sharing,
intragroup conflict, controversy and process facilitation in determining group
performance in a complex, ill-structured task. Second, this study will provide a further
assessment of the relative eflicaCy of structured conflict methods and the consensus-

seeking approaches to group decision making. Third, it attempts to replicate the findings

67
of Stasser with regard to the failure of decision making groups to identify "hidden
profiles," extending our knowledge of that issue to ill-structured situations with
meaningful measures of group performance. Finally, this study will provide a preliminary
test of the relative merits of a modified form of DI hypothesized to result in more
information sharing and lessened intragroup conflict. Ultimately, this study will help to
provide a better understanding of how structural interventions can be designed to improve
both individual and group outcomes in decision making settings via the promotion of

information sharing and the reduction of intragroup conflict.

METHOD

Participants

Research participants in this study consisted of 240 college students enrolled in
one of several tmdergraduate psychology courses offered at a large public university in
the midwest. Individuals participated in this study as part of four-person groups and data
was collected on 60 groups in total. Of the 60 groups, 14 were composed entirely of
females, 17 groups had three females and one male, 18 groups were composed of two
females and two males, eight groups were made up of three males and one female, and
only three groups consisted entirely of males.

Individuals who took part in the study received course credit or extra credit for
participating in the study. In addition, a financial incentive was employed to maximize
participant motivation in the study. In each of the three study conditions described below,
top performing teams were awarded $80 ($20/person), second-place teams were awarded
$60 ($15/person), and third-place teams were awarded $40 ($10/person).

W

The task employed in this study simulated the operations of a hypothetical
regional U.S. airline organization, "SouthEast Airlines" (Devine, 1995). The simulation
is intended to be a moderately realistic strategic decision making task conducted in a face-
to-face setting. In "SouthEast Airlines," groups of four individuals represent the top
management team of "SouthEast Airlines" charged with the task of creating a strategic
business plan for an upcoming period of airline operations. Each member of the group is

assigned a position as an executive vice-president in the company responsible for one and

68

69
only one of the following areas: Flight Operations, Industry Analysis, Marketing, or
Finance. The overall object of the task is to formulate a plan resulting in maximum profit
for the organization. In order to create such a plan, each group must make numerous
interdependent decisions involving choices about such things as service routes, aircraft
route assignments, facility locations, fare prices, advertising media and spending levels,
as well as aircraft sales and purchases.

The simulation consists of two major parts: (1) an Individual Preparation phase in
which participants are allowed to study the information provided to them and (2) a Group
Discussion Phase in which group members are allowed to interact in the process of
forming their collective strategic plan. (See the procedure section below for further
details on experimental protocol.) Group performance in the simulation is defined as
profit earned by the group's plan when resolved according to the simulation's algorithms.
Appendix A contains all task materials associated with the "SouthEast Airlines"
simulation.

"8th Airlines" is intended to represent real-world situations where
information relevant to group decision making is distributed across a number of
"experts." In "SouthEast Airlines," much of the information available to the group is
distributed across the four vice president positions mentioned above, each of which
represents a difl‘erent fimctional area of the organization. Prior to beginning the
simulation, after having been assigned to one of the four vice president positions, each
group member receives a packet of information corresponding to the particular position.
The information contained in each position packet can be divided into two parts:
information provided to ALL group members and information provided to only one
member (although a few pieces of information were provided to two players out of logical
necessity). All group members received a document entitled "8th Airlines' Year-

End Report," a document containing information about the company's operations in the

70
last fiscal year. Each vice-president also received a packet of information ("Memo") from
his or her respective staff (see Appendix A also for these documents). The material
contained in each vice-president's memo concerned information relevant to that particular
frmction within the organization. Table 2 provides a summary of the information

provided to the four vice-presidents in "SouthEast Airlines."

71

Table 2.

Information Provided to Vice—President Positions in "SouthEast Airlines."

 

Document-Recipient Information Concerning:

 

Year-End Report Last year's:

(ALL)
Operational routes
Aircraft assignments
Route Market Share
Fare prices

Revenue & Costs (total and by route)
Route Return on Investment (ROI)

Memo to Last year's fuel costs (by route)

VP Flight Operations Aircraft operating characteristics

Eflect of #Daily Flights, Aircraft
Accommodations, and #Flight Stafi
on Market Share

Memo to Potential expansion routes

VP Industry Analysis Expected Competition levels
Expected Passenger Demand values
Round-trip Distances

Industry averages for Flight Staff

 

 

 

 

Memo to Setting optimum fare prices

VP Marketing Advertising costs
Advertising effects
Advertising media

Memo to Personnel cost projections

VP Finance Facility cost projections

Existing loans
Aircraft sales

 

 

 

 

72

As can be seen from Table 2, each role in the simulation received basic
information describing operations during the last fiscal year plus some important
information which no other participant received. Given the sub-optimal quality of the
existing strategic plan, each vice president was in a position to recommend some changes
that would improve upon existing operations. However, although possible to improve
profit by simply employing unilateral suggestions from each vice-president, the
simulation was designed so that each group, in attempting to integrate the
recommendations of each position, would have to resolve the dilemma of "expansion"
versus "consolidation."

In order to allow for the examination of dialectical inquiry with "diametrically
opposed" plans, information was provided to the Vice-Presidents of Flight Operations
and Industry Analysis leading them to adopt one or the other of these two strategies. In
particular, information provided to the Vice-President of Flight Operations supported a
"consolidation" strategy based on dropping high-cost existing routes, re-allocating
aircraft, increasing daily flights and flight stafi‘, and buying a new fuel-efficient aircraft.
The Vice-President of Industry Analysis, on the other hand, received information
suggesting the need for an "expansion" strategy centered on dropping most of the existing
routes, adding many new routes with limited daily flights to each city, and buying a
number of large, expensive new aircraft. Information provided to the Vice-Presidents of
Marketing and Finance was theoretically "neutral" with regard to each of these strategies.

However, since no one other than the Vice President of Industry Analysis had
information on new routes, the situation in fact tended to begin as a "three-against-one"
coalition in favor of an efficiency-consolidation approach.

To reinforce the underlying dialectic in the simulation, certain types of
information were provided in a probabilistic fashion. In particular, the information

provided to the Vice-President of Industry Analysis concerning the competition level on

73
various routes was given as a percentage likelihood, while passenger demand values were
provided in terms of a range. This uncertainty was designed to add realism to the task in
that outcomes could not be perfectly determined in advance, as well as allow assumptions
to enter into discussion as participants were forced to "fix" random values in their own
minds in order to create a cohesive plan.

In summary, "SouthEast Airlines" is intended to be a low-fidelity simulation
involving: (1) uncertainty, (2) heavy information processing demands and (3) the
existence of multiple perspectives and approaches with regard to how the group can best
satisfy its objective. The simulation incorporates a number of concepts relevant to
strategic management, including price-demand relationships, competition, environmental
tmcertainty, local monopoly, market share, advertising, operating costs. Information cues
in "8th Airlines" were distributed so that no group member had access to all (or
even the majority) of the available information. Some information was known to only
one member, while most of the information was known to all members before group
discussion. However, all rules and information necessary for creation of a high-quality
plan was provided. Thus, as in Stasser's previous work on information sampling, groups
were in the position of having to "uncover" the knowledge possessed by all members in
order to resolve the various trade-offs that arise in the course of specifying operations.
Research Design

This study involved one manipulated between-groups factor, Inquiry Method,
composed of three levels: (1) Consensus-Seeking (CS), (2) Traditional Dialectical
Inquiry (TDI), and Synthesis Dialectical Inquiry (SDI). Groups were randomly assigned
to one level of the Inquiry Method factor with the constraint that all conditions must have
run once before any condition could repeat. This established a repeating cycle of three-
group sequences where each condition appeared once in the sequence in a random order.

Eighteen full sequences were conducted (54 groups, 18 of each condition), and then a

74
selected number of each condition were run in a randomly determined order in an attempt
to arrive at equal numbers of each condition after some groups were removed due to the
loss of data or invalid group plans. In total, 19 CS groups, 20 TDI groups and 21 SDI
groups were run, although some groups were dropped from one or more analyses due to
missing data (see section on missing data in the Results for further details).

Inquiry Method was manipulated by providing videotaped instructions to all
groups concerning the procedures to be used in completing their task. Instructions were
presented to groups immediately before the Individual Preparation phase, and a short
verbal reminder was provided immediately prior to Group Discussion. A printed copy of
the group's inquiry method instructions was also provided in the role packet of
information given to each participant so that instructions would be available at all times.

The CS condition was intended to provide a baseline condition with respect to the
two DI conditions. Appendix B displays the instructions presented by videotape and in
writing to participants in groups using the consensus-seeking method. As evident in the
instructions, groups were given general instructions to present and consider all views,
manage conflict productively, and avoid adopting a plan immediately if everyone seems
willing to accept it. Groups were instructed to discuss ideas until all members were
willing to accept the features of a particular plan, at which point the group was said to
have reached consensus. No special role assignments were made in the CS condition and
group discussions were not constrained to begin in any particular fashion.

The Traditional DI (TDI) condition was intended to represent dialectical inquiry
in a fashion similar to the way it has been operationalized in the literature - with explicit
thesis and antithesis assignments, but no assignment of the synthesis function. Appendix
C presents the instructions given to groups employing the Traditional DI method.
Inspection of this appendix shoWs that TDI groups were instructed to begin Group
Discussion with the following sequence of events: (1) The VP of Flight Operations

75
presents Plan A, (2) The VP of Industry Analysis presents Plan B, (3) The VP of Industry
Analysis critiques Plan A, and (4) The VP of Flight Operations critiques Plan B. The
dialectic process was to occur in the first 20 minutes of Group Discussion, after which
general discussion among all members was allowed to begin and continue until a plan had
been reached which was acceptable to all group members. In the TDI condition, the
Vice-Presidents of Marketing and Finance were given no specific role and were asked to
hold their questions and comments until after the dialectical process between the other
two vice-presidents was finished.

As noted above, information provided to the two vice-presidents involved in the
creation, presentation and critique of plans was explicitly designed to yield two plans that
were "diametrically opposed" to one another. These two individuals also received a
supplementary set of instructions in order to help them enact their role in the dialectical
process. See Appendix D for the role instructions given to the two vice-presidents
involved in the dialectical process.

The Synthesis DI condition was designed to employ dialectical inquiry in a
manner consistent with that originally discussed by Mason (1969) and Mason and Mitrotf
(1981) by explicitly assigning the synthesis role to members of the group. Appendix E
depicts the instructions given to groups in the SDI condition. As can be seen from the
appendix, the SDI condition was implemented in exactly the same fashion as the TDI
condition with one exception: The Vice-Presidents of Marketing and Finance were
instructed to take a synthesis role after the 20 minute dialectical process involving the
other two vice-presidents had ended. Specifically, these two vice-presidents were asked
to summarize what had taken place in the "debate," ask questions about unclear
recommendations, integrate ideas where possible and provide some structure for the
remainder of the group's discussion. Similar to the role instructions provided to the

dialectic presenters/critiques, the vice-presidents asked to play a synthesis role were

76
given detailed instructions in order to help them fulfill their role in the group's discussion.
See Appendix F for a copy of the synthesis role instructions.

Table 3 provides a summary of the role assignments in each study condition.

Table 3.

Role Assignments by Study Condition.

77

 

 

 

 

 

 

 

 

 

Role Consensus- Traditional DI Synthesis DI
Seeking

VP Flight None Create/Present Plan Create/Present Plan

Operations (Consolidation) (Consolidation)

VP Industry None Create/Present Plan Create/Present Plan

Analysis (Expansion) (Expansion)

VP Marketing None None Synthesis role

VP Finance None None Synthesis role

 

 

78

Two manipulation checks were conducted to assess the degree to which groups
effectively employed their assigned inquiry method. The first check involved several
items in the post-experimental questionnaire asking respondents about the quality of role
performances in the group. See Appendix N for the post-experimental questionnaire
measure. The second check involved a simple dichotomous judgment regarding whether
a group should be removed from the analysis made by a single rater based on the group’s
videotaped interaction and discussion. The data from these two checks were used to
identify groups which did not take the task seriously and/or failed to learn the task well
enough to make their results meaningful.

Procedure

Groups for this study were formed using standard sign-up sheets circulated to
undergraduate psychology courses. Individuals were scheduled to arrive at the lab in
groups of six, of which four individuals were needed to form a group for the "SouthEast
Airlines" simulation. In the event that more than four persons showed up for a session,
the first four individuals were formed into a group and "extra" individuals were moved to
another room where they participated in a study unrelated to the one described here.

The study was conducted in four phases: Pre-Experimental, Individual
Preparation, Group Discussion, and Post-Experimental. The Pre-Experimental phase
lasted approximately 20 minutes, the Individual Preparation phase lasted for 60 minutes,
groups were allowed up to 75 minutes in the Group Discussion phase, and the Post-
Experimental phase lasted about 10 minutes. Table 4 provides a chronological listing of
the activities in this study broken down according to these four phases. The remainder of

this section describes each of these phases in more detail.

79

Table 4.

Sequence of Events in "SouthEast Airlines."

Pre-Experimental phase

1. Participants arrive & receive "Overview"

2. Wonderlic test administered

3. Individuals randomly assigned to vice—president positions
4. Individuals receive position-specific packets

a. "Year-End Report"

b. Position-specific "Memo"

c. Task knowledge measure

(1. Inquiry Method instructions

e. Role instructions (in TDI and SDI conditions)

5. Groups watch tape with Inquiry Method instructions

Individug PrepaLation Phfie
6. Individuals allowed 60 minutes to prepare for Group Discussion

a. In TDI and SDI conditions, VPs of Flight Operations and Industry Analysis
prepm plans ‘

b. In SDI conditions, VPs of Marketing and Finance review instructions on
how to fulfill synthesis role

Group Discussion phase

7. Groups receive inquiry method reminder
8. Groups allowed 75 minutes to fill out planning document

a. CS groups have 75 minutes of unstructured discussion
b. SDI and TDI groups complete dialectical process in first 20 minutes, then

Post-Exmgrp’ ental phase

9. Groups complete post-experimental questionnaire
10. Groups debriefed and dismissed

80

The pre-experimental portion of the study began when four individuals had
arrived at the experimental room and were seated around a large rectangular table. As
they arrived, group members were given a written document providing an overview of the
experiment, identifying their goal as a group, and describing the simulation environment
(including the general algorithms for figuring revenue and costs). Sessions were started
five minutes after the fourth group member arrived, ensuring that all participants had at
least five minutes to look over their introduction (determined in pilot testing to be
sumcient time to read through the document at a comfortable pace). Five minutes after
the arrival of the last individual, the Wonderlic Personnel test was administered to the
four group members. When the test was finished, group members were randomly
assigned to a vice president position in the group and moved to a specific location at the
table so that the respective positions were always in the same seats, providing a
standardized spatial arrangement and making it easier for observers watching the
videotapes.

Once group members were seated in their appropriate spots, the experimenter
passed out a packet of information to each member of the group and set out name plates
that identified each group member's position title. Each group member received a packet
of information containing: (1) a year-end report, (2) role-specific information relevant to
future planning, (3) a task knowledge measure, (4) a hard copy of the group's inquiry
method instructions, and, in the two DI conditions, (5) specific instructions relevant to
performing their assigned role during group discussion (i.e., presenting and critiquing
plans, enacting a synthesis role).

After distributing the position packets, the experimenter reviewed the purpose of
the task and the ftmction of each component in the information packets provided. The
experimenter then asked if there were any questions and proceeded to play a short

videotape detailing the instructions for the inquiry method to be used by the group. The

8 1
videotape instructions were identical to the instructions given to each member earlier in
his or her role packet and participants were instructed to read along on their hard copy as
they listened to/looked at the videotape. The video provided instructions concerning how
the group should structure its activities during group discussion and, for the DI
conditions, identified the roles each group member would have and the sequence of
activities that should be followed once the Discussion Phase began.

As noted earlier, in both the Traditional and Synthesis DI conditions, participants
playing the Vice-Presidents of Flight Operations and Industry Analysis were asked to
create their own individual plans and present them to the group at the start of group
discussion. These individuals were also instructed to provide a critique of the "opposing"
plan after the two plans had been presented. Additionally, in the Synthesis DI condition,
the vice presidents of Marketing and Finance were instructed to act in a facilitating role
and were given instructions regarding this role.

After the videotaped instructions related to inquiry method were played,
participants were again provided with an opportunity to ask questions. After answering
any questions, the experimenter set a timer for 60 minutes and then left the room,
returning every so often to check on the group and answer questions. This marked the
beginning of the Individual Preparation phase.

In the Individual Preparation phase, group members were allowed 60 minutes to
review the materials provided to them and prepare for the upcoming Group Discussion
phase. During the Individual Preparation phase, group members were provided with
scratch paper and calculators and were allowed to write on the materials they had been
given. Individuals were instructed to use the time allowed to prepare for the upcoming
group discussion subject and prepare for their assigned role during the Group Discussion

phase. Group members were also instructed to complete the task knowledge measure

82
before the end of the Individual Preparation phase, and were given several reminders to
do so during the course of the phase.

After the 60 minutes allocated for Individual Preparation had expired, the Group
Discussion Phase began. The experimenter re-entered the room, gave a brief reminder of
the instructions the group were given before the Individual Preparation phase, and started
a video camera set up to film the group's discussion. The final action taken by the
experimenter before leaving the main room again was to start a clock timer set for 75
minutes and turn it so that all group members could see its face. Each group was allowed
75 minutes to arrive at a collective group strategic plan covering desired operational
activities. Groups were instructed to complete their strategic plan using the appropriate
form and return it to the experimenter when he or she returned at the end of the Group
Discussion phase.

Again, Consensus-Seeking groups were given general instructions to explore all
options, press for clarification, avoid win-lose statements and refrain from conflict-
reducing "tricks." In the DI conditions, groups utilized a format wherein the two plans
created by individual members will be presented and debated. In the T'DI condition, the
two group members not involved in the creation of the plan or counterplan were given no
special instructions. SDI groups received the same instructions as the TDI groups with
one exception: after both plan and counterplan have been presented, the two group
members not involved in their creation were to summarize the major points of each plan,
ask questions to clarify confusing statements or recommendations, ofl‘er integrative or
compromise plans, and otherwise provide a structure for the remainder of the group's
discussion. During the course of the 75 minute Group Discussion phase, the

experimenter checked at the door to monitor progress toward completion and insure that

groups stayed on task.

83

When groups were finished (or after the expiration of the 75 minutes allowed), the
experimenter re-entered the room, collected the group's plan and administered the post-
experimental questionnaire. Upon completing this questionnaire, participants were
thanked for their participation, told they would be contacted in the event of their winning
a prize for superior incentive, and given a debriefing sheet. At this point, participants
were invited to ask any further questions and then dismissed.
Measures

As the primary focus of the study, one outcome and four process variables were
measured in this study at the group level: (I) Group Performance, (2) Information
Sharing, (3)1ntragroup Conflict, (4) Process Facilitation and (5) Controversy. In addition
to these variables, three group composition variables were also measured and used as
control variables in the analyses: cognitive ability, task knowledge, and sex composition.
In this section, I describe each construct, identify its dimensions, and discuss how a
measure of each was generated. In the next section, a discussion of rater training is
offered. Because of complications in the estimation of measurement reliability,

information on this is reserved for the Results.
Group Performance represents the degree to which the group was able to

accomplish its primary goal of creating a plan which would bring in more profit than the
previous fiscal year. Group Performance was determined in a nonj udgrnental fashion by
using a set of algorithms to determine the profit that would have been generated for each
company in the simulation environment. More specifically, the decisions that groups
make about such things as the routes to be serviced, the type of plane assigned to each
route, the number of daily flights, etc., were translated into revenue and costs for each
group. Group Performance was determined by calculating total assets (i.e., cash plus
revenue) and subtracting total liabilities (costs plus debts). All group members were

provided with the basic terms and formulas used to determine revenue and costs in the

84
"Overview" materials before the beginning of the study. Since the strategic plan created
by the group represents the entire group regardless of the distribution of effort expended
by individual members, Group Performance was treated as a group-level outcome
variable.

Process Variables. Information Sharing refers to the act of making an information
one available to all group members for use in group-level planning. In SouthEast
Airlines, information cues are bits of factual information provided to group members
concerning: (1) the reported level or status of variables during the last year of operation,
(2) the projected level/status for variables in the upcoming (i.e., simulated) year of
operation, and (3) the relationships between variables. Examples of each type of
information cue include (1) the average fare charged by the competition on a given route
last year, (2) the expected passenger demand for a particular route next year, and (3) the
formula for route revenue, respectively.

At the beginning of the simulation, some of the information cues in the simulation
were known to all group members, a few to two or more group members, and most to
only one group member. Items available to all members at the beginning of the
simulation are termed common, while those known only to one member (or in a few
cases, two members) are designated as unique. An instance of information sharing was
considered to have occurred when an information cue known to only a subset of group
members was spoken aloud dming group discussion. Information cues known to two
members before discussion were considered to be “unique” as this seemed closest to the

spirit of Stasser’s earlier work where “shared” (i.e., “common”) information was known
to all group members before discussion. Of the 700 information cues available in the
simulation, only 12 cues (<2%) fell into the category of being known by two members

before discussion.

85

Information Sharing measures were obtained from video tapes of group
discussions which allowed the coding of video and audio behaviors of group members. A
measure of Information Sharing was calculated by having a trained observer count the
number of unique information cues shared during group discussion (see discussion of
rater training to follow). In other words, information cues available to only one (or two)
group members before discussion which were spoken aloud were counted. Videotapes
were scored by one rater using a checklist of the information cues provided to groups in
terms of the number of individual-level cues shared by each group member. Appendix G
contains the checklist used by raters which contains all information cues presented to
groups in the simulation. Each group’s Information Sharing score was determined by
adding up the number of unique information cues shared during discussion and using this

total score to represent the group in a fashion similar to Stasser et al. (1989).
Intragroup conflict is a multidimensional construct characterized by negative

affect, hostile interpersonal behaviors and distorted cognition within the group. Although
conflict tends to occur between identifiable subsets of group members, the literature
suggests that conflict afl‘ects the entire group. For instance, if two members are at each
other’s throats, this can affect the entire group by (1) directly consuming scarce group
resources (e.g., time) and (2) arousing negative affect in those members who are
“watching.” Group members who are watching may become tense or anxious, frustrated
at the time wasted in the dispute, and angry at those arguing. This is consistent with
Pruitt and Rubin (1986) who noted that conflict tends to spread out over all group
members, lose its focus, encompass broader issues and become more general.

Intragroup conflict was measured in two ways: (1) with a 10-item questionnaire
using a five-point Likert response format, and (2) by having trained raters watch the
videotape of each group during discussion and rate the degree of intragroup conflict

exhibited by each group. Appendix H contains the instrument used by raters to provide

86
judgments of intragroup conflict. The lO-item Intragroup Conflict questionnaire measure
is presented in Appendix I. An internal consistency reliability estimate was performed on
responses to these 10 items and two items displaying a negative item-total correlation
were removed. (These Items are marked with an asterisk in Appendix I). The coefficient
alpha reliability estimate for the final eight-item scale was .78. Individual scale scores on
the questionnaire measure of conflict were then created by summing the unweighted
responses on the final eight items and a score was then assigned to the group by averaging
the scores of the four individual group members. (See the Results for further discussion
concerning the meaningfulness of this procedure.)

Process Facilitation is the degree to which the individuals in a group engage in

behaviors serving to facilitate the processing of information during group decision
making. Process facilitation behaviors should aid groups in identifying and using the
informational resources at their disposal. Four general sets of behaviors can be
distinguished in the literature: (1) Summarizing/reflecting, (2) Asking
questions/clarifying, (3) Integrating and (4) Focusing/structuring. Although behaviors are
engaged in by individual group members, they benefit the entire group. Therefore, by
definition, process facilitation is something done by individuals in the interests of the
group. The instrument used by raters to provide ratings of process facilitation on each of
its four component dimensions is included in Appendix J. An overall score on the
measure was generated for each group by summing the unweighted observer ratings for
the four component dimensions.

Controversy represents the degree to which individuals in a group hold different
ideas, opinions, theories, and viewpoints with regard to how the group's goal can best be
achieved. In a practical sense, controversy can be seen as the extent to which group
members (1) indirectly challenge views or positions taken by other members of the group,

(2) explicitly disagree with views or positions expressed by other members concerning

87
how things should be done and (3) present multiple options, alternatives and plans for
accomplishing the group's goal. Furthermore, although past research has tended to view
controversy as a construct that exists within an individual decision maker, it is treated
here as a group-level construct involving the identification of multiple "pa " to the
group goal. The instrument used by raters to provide judgments of controversy within
groups is contained in Appendix K. A group-level index of controversy was then formed
by summing the unweighted observer ratings for each of these three component
dimensions.

Because of the poor reliability associated with this measure of controversy (to be
addressed in the Results section), a second measure of controversy was generated using
new raters and a modified ratings instrument. In an attempt to improve reliability, a post
hoc effort was made to identify the specific types of substantive controversy which might
occur in "SouthEast Airlines" and the videotapes of group discussion were recoded by the
raters. Appendix L contains this alternative measure of controversy. As can be seen, two
types of controversy were distinguished in this measure: (1) task-related controversy (or
strategy-related issues) and (2) process-related controversy (structure and agenda-related
issues). The various dimensions of controversy were collectively generated by the
experimenter and the two research assistants who provided the original controversy
ratings. As with the original controversy scale, group-level scores on the revised measure
of controversy were generated by summing the unweighted ratings on each of the nine
dimensions. See the section on "Quality of Measures and Manipulations" in Results for
further details.

Comppsition Variables. In addition to the process variables described above,
three group composition variables were also measured - cognitive ability, task
knowledge and sex/ gender ratio. These variables are “composition” variables in the sense

that, for each variable, the score assigned to the group is a perfect ftmction of the four

88
individual values. Although there are certainly many dimensions along which the
members of a group may differ, cognitive ability and gender were included because of
empirical evidence indicating their effects on group process and/or performance (e.g.,
Tziner & Eden, 1985; Wood, 1985; Sundstrom & F utrell, 1993; Milliken & Martins,
1996). Task knowledge was included on the basis of its logical relationship with group
performance. In contrast to the group structure variable (i.e., inquiry method), these three
composition variables were not manipulated but simply measured for use as control
variables or examination in supplementary analyses.

Cognitive Ability refers to the degree to which the members of a group are able to

learn new tasks, adapt to novel situations and problem-solve. In complex strategic
decision making tasks, it is reasonable to assume that the cognitive ability of all members
is important. Each member has the “duty” of recognizing, comprehending and sharing
important information that he or she possesses. After information is “out on the table”
for all group members, each group member can add to, or withhold fiom, the collective
product being fashioned. As a result, because of the additive nature of these processes, it
seems appropriate to generate a score for each group based on the average individual
cognitive ability of each of its members, although it certainly would have been possible to
argue for and employ other combination rules. Ultimately, given the dearth of research
on the topic, the decision to use the averaging function was somewhat arbitrary.
Individual Cognitive Ability was measured by giving each member of the group a
short, paper-and-pencil measure of cognitive ability, the Wonderlic Personnel Test. (This
measure is not shown due to the proprietary nature of the instrument). The "Wonderlic
Personnel Test & Scholastic Level Exam: User's Manual" (1992) notes that the
Wonderlic measures of "g," or general intelligence, and demonstrates excellent reliability

as well (with various internal consistency and test-retest estimates ranging from .84 to

89
.92). Scores for individual members were averaged to create a mean score representing
the entire group.
Task Knowledge was defined as the amount of role-specific knowledge learned by

the members of each group. Because of the design of the task, it was possible to
distinguish between two types of task-related knowledge: general and position-specific.
General knowledge questions pertained to the basic "rules" of the simulation and were the
same for all group members. Position-specific knowledge corresponds to unique
information known only to a subset of group members (usually one). With regard to the
most appropriate function relating individual task knowledge and the knowledge
possessed by the group as a whole, again an analysis of the task seems to suggest that a
score assigned to the group should weight equally the information of each member in an
additive fashion (i.e., an average). All members possess important information that could
be used to improve the group product if shared with other members and, at least by the
intention of the design, no member possessed information that could be considered more
crucial than any other group member’s information. As a result, a score characteristic of
the group was assigned on the basis of the arithmetic average of the four individual
members, but this is again acknowledged to be somewhat arbitrary.

A measure of general task knowledge and four measures of position-specific
knowledge were created, and each group member received a paper-and-pencil test of the
general knowledge items as well as the appropriate position-specific items during the
Individual Preparation phase of the experiment (see below). Appendix M contains the
general measure of task knowledge given to all positions, as well as the four position-
specific measures. For each group, task knowledge was measured by summing the four
position-specific knowledge scores for the four individuals in the group.

Sex/gender ratio refers to the relative number of men and women within a group.

This variable was included in the study on the basis of past research indicating

90
performance differences in small groups as a function of the sex of group members. In
particular, Wood (1985) conducted a meta-analysis examining studies of small group
performance in the context of single-sex groups. Across all studies, Wood found a
tendency for male groups to perform somewhat better than female groups but also
identified task type as a potential moderator of the sex composition-performance
relationship. More specifically, Wood’s analysis suggested that male groups perform
better on agentic tasks (i.e., those involved with giving opinions and recommendations)
while female groups do better on communal tasks (i.e., those involved with fostering
friendship and agreement). Given the clearly agentic nature of the present task, it seems
reasonable to expect that the sex composition of the groups will affect group
performance. However, given that this finding is based on research utilizing single-sex
groups and few or no process measures of interest here, it is not clear how mixed-sex
composition should affect group process or performance. As a result, no specific
hypotheses regarding sex composition seem appropriate.

Sex composition was operationalized as the number of men in each group of four
individuals (0-4). Although it would have been possible to generate a sex/ gender
composition score in some fashion other than this, parsimony suggests the use of simpler
representations until the utility of more-complex representations has been established. As
such, although again somewhat arbitrary, the decision was made to represent sex/ gender
composition as a simple cormt variable.

Rater Training

Overview. Between the conclusion of pilot testing in the fall of 1995 and the
beginning of actual data collection for this study in spring 1996, a number of research
assistants were recruited and trained as observers of group discussion. Initially, seven
research assistants were trained to observe videotapes of group discussion and either code

instances of information sharing or provide ratings on one of the three judgmental

91
measures (i.e., intragroup conflict, controversy, and process facilitation). Each research
assistant was trained to observe only one construct (i.e., intragroup conflict or
information sharing or controversy). As a result, there was no possibility of method bias
in the relationships among process facilitation, controversy, intragroup conflict and
information sharing that would have resulted if one rater had provided multiple measures
for each group.

To begin with, one research assistant was assigned to code information sharing
and two were assigned to each of the judgmental measures. It was anticipated that the
one research assistant would provide all the ratings for information sharing, while the two
raters assigned to each of the judgmental measures would rate half of the groups, plus
five groups rated by the other rater. This was done in order to provide a sample of 10
groups redtmdantly coded for each of the three judgmental measures which would allow
for an estimate of interrater reliability.

Gron Training. The training procedure used for all four measures was as
follows. The seven research assistants were gathered for a preliminary meeting where
they received an overview of the task, construct definitions, and a copy of the respective
rating form (or checklist) they would be using. Focusing on each construct in turn, the
experimenter went over the definitions and behavioral examples of each sub—dimension.
After this, a period of time was devoted to the issue of what defines an "instance" of some
dimension. In essence, an "instance" was defined as one of two types of activity
matching the definition of a component dimension: (1) a discrete action or utterance on
the part of an individual or (2) a sequence of verbal interaction with an identifiable
beginning and end. After this, raters were given instructions for how to go about making
their ratings (see next section for details). Finally, raters viewed several segments of
videotape from pilot groups in order to practice identifying behaviors that corresponded

to the various dimensions of each construct. At this point, formal training ended, but the

92
experimenter met periodically with each rater in order to gauge drift from the standards
set during training. For practical considerations, raters were allowed to watch the
videotapes of group discussion and complete their ratings (or fill out the information
sharing checklist) using a VCR at their home.

_R_ater igstructions, In the case of Information Sharing, coders were instructed to
listen carefully to each group's dialogue and record all quantitative information
represented in the simulation spoken by any of the group members by circling that piece
of information on the Information Sharing checklist. Coders were instructed to record
questionable instances of information sharing on the last page of the check-list, as well as
any item of shared information which they considered to be meaningful that was not
represented in the checklist. In practice, this portion of the checklist was rarely used.

For the three judgmental measures, raters were instructed to watch the entire tape
of group discussion and record each "instance" of conflict they observed using tic marks
in the margin to the right of each dimension rating. Raters were told to use the total
number of tie marks tallied for each dimension as a guide in making their ratings for each
dimension. Because it was expected that there would be a fair amount of disagreement at
the "micro" level of behavior, it was stressed that raters should make their ratings for each
dimension on the basis of their overall impressions and use the tic marks for guidance.
As such, no explicit quantitative guidelines were given beyond those implied in the
anchors for each scale. Again, the ratings provided on each dimension were summed to
yield an overall conflict score for each group.

Due to unexpected turnover in research assistants over the course of actual data
collection and the slow, uneven pace of coding, three additional raters were trained to
observe and rate groups on intragroup conflict and one additional rater was trained to
observe and rate process facilitation. As a result, five research assistants ended up

making ratings of intragroup conflict, three research assistants made ratings of process

93
facilitation and two research assistants rated controversy. "Second-wave" research
assistants were trained in the same manner as the original research assistants with two
exceptions: (1) the initial training session was conducted individually for each new
assistant and (2) instead of watching behavioral examples as a group, new assistants were
given several "practice" videotapes consisting of pilot groups rtm during the fall. Each
new research assistant then met individually with the experimenter to discuss the practice

ratings and receive feedback.

RESULTS

Overview of Results Section

The results of the analyses performed for this study are divided into several parts.
First, I discuss the cell sizes for the various conditions of the study and the loss of groups
due to attrition. Second, in the section entitled "Quality of Manipulations and Measures,"
there is a discussion of pilot testing, the reliability of measured variables used in the
study, several analyses best viewed as manipulation checks and an examination of the
effects of several potential nuisance variables. In the following section, the results of
analyses related to the efi‘ects of various group processes on other group processes
(Hypotheses 1-5) are reported. In the fourth section, the effects of the three inquiry
method conditions on group performance (Hypotheses 6-8) are considered. Finally, in the
fifth section, the results of several exploratory analyses are reported.
Gropp Attrition

In total, the CS condition was run 19 times, the TDI condition 20 times, and the
SDI condition 21 times. However, one or more process-related measures was lost for six
groups (e.g., videotape failure during group discussion), and eight groups failed to submit
a valid business plan according to the rules of the simulation. In addition to these 14
groups, one group was removed from the analysis because its score on Information
Sharing was four standard deviations above the mean for all groups. Table 5 provides
details on the groups that failed to provide acceptable data on one or more study

94

95
measures. Unfortunately, these two categories were mutually exclusive, resulting in the
maximum possible loss of 15 groups for correlations with strategic planning quality. On
the plus side, the groups lost were distributed fairly evenly across experimental
conditions. As a result, there were 53 groups with complete data for the analyses which
did not involve group performance (1 9, l7, 17 groups for CS, TDI and SDI conditions,
respectively) and 45 groups with complete process and outcome data available for use in
the regression analyses involving group performance (16, 15, 14 groups for CS, TDI and

SDI conditions, respectively).

Table 5.

96

Cell Sizes for Study Conditions and Statistical Analyses.

 

 

 

 

 

 

 

 

 

 

 

 

 

Consensus Traditional DI Synthesis DI

Total Sessions Run 19 20 21
Groups missing one or more 0 2al 4b
process measures
Groups which failed to submit 3 2 3
a valid business plan
Groups removed as outliers 0 1° 0
Final N for analyses NOT 19 17 17
involving Group Performance
Final N for analyses involving 16 15 14
Group Performance

Notes

‘One group was inadvertently left out of the controversy recoding; another group was
unable to be recoded for controversy due to tape breakage

bOne group was inadvertently left out of the controversy recoding; another group failed to
complete the post-experimental measure containing the conflict scale; process measures

were rmavailable for two additional groups due to technical failures during videotaping.

°One group was four standard deviations above the mean on Information Sharing

 

97
Qu_alitv of Mamulptions and Measures

Pilot Testing. Pilot testing for this study was conducted in the spring of 1995 and
then again in the fall of the same year. Six groups were run in the spring with the
experimenter present taking notes during group discussion. Another 12 groups were run
in the fall, with revision and development after the first block of six had been completed.
The final six pilot groups were videotaped using standardardized experimental
procedures. Using these last six groups, an initial estimate of inter-rater reliability was
calculated for information sharing, intragroup conflict, controversy and process
facilitation using this sample of six groups.

Because only two research assistants were available during pilot testing, it was
necessary to train these two individuals to make ratings for each of the four measures.
Note that in the actual study, each rater was trained to rate (and only rated) one measure.
Observers were instructed to keep the constructs distinct in their minds, avoid "halo" as
much as possible, and review the tape as needed after watching it once. Each observer
watched the first 15 minutes of each group discussion and made ratings for each of the
four measures noted above. Then, for each of the six groups, total scores generated by
Rater 1 on the four measures were correlated with the total scores generated by Rater 2.

The four reliability estimates generated in this fashion for the six pilot groups
varied widely in magnitude. Consistent with earlier work by Stasser suggesting that
information sharing could be reliably measured from audio tapes, the interrater reliability
correlation for information sharing was extremely high (50, = .97). For intragroup conflict
and controversy, interrater correlations were a little lower but still quite acceptable (rxx =
.85 for controversy; 3;,“ = .75 for intragroup conflict). On the other hand, the estimate of
interrater reliability was calculated for process facilitation using a sample of six groups
that were observed and rated by two independent raters. The resulting interrater

correlation was rm = -.17. As a result, rater instructions and definitions of the process

98
facilitation dimensions were revised and more behavioral indicators were generated
before this rating instrument was used in the actual study.

RLligbility of Measured Variables. For the actual study, reliability estimates for
the intragroup conflict, controversy and process facilitation measures were calculated
again using a difl‘erent procedure than in the pilot study. The reliability of the
information sharing index was not re-estimated given its nonj udgmental nature and the
extremely high estimate of interrater reliability obtained during pilot testing. Also, as the
group performance score was derived from computational formulas involving only basic
addition and multiplication, the reliability of this variable was not estimated and assumed
to be near 1.0.

As noted earlier, measures of intragroup conflict, controversy and process
facilitation for the actual study were obtained by having a single trained observer watch
each videotape of group discussion and provide ratings on the particular measure for
which he/ she was trained (see Method for details of training). As a result, each videotape
of group discussion was observed by four difl‘erent individuals, each trained to provide
ratings on one specific measure.

To estimate the interrater reliability for intragroup conflict, controversy and
process facilitation, a sample of videotapes was rated a second time by a second rater
trained to rate that particular dimension. Interrater reliability estimates were obtained for
the three judgmental measures by calculating the Pearson correlation for the groups coded
by two independent raters. Table 6 presents information related to the calculation of
inter-rater reliability correlations for intragroup conflict, controversy and process
facilitation. As can be seen in the table, approximately 20% of the groups were
rodlmdantly coded by two raters on each of the three measures.

99
Table 6.

Interrater Reliability Estimates for Process Measures.

 

 

 

 

 

Observer N Raters Initial Final
Rating rxx rn**
Measure

___
Intragroup 6 2 & 3 .84 ??
Conflict

10 4 & 5 .10

Controversy 12 7 & 8 .23 .56‘
Process 12 9 & 1o .47 .84"
Facilitation

 

 

 

 

 

 

 

** After post-hoc scale modification

'Reliability estimate for revised measure of controversy using revised scale and sample of
each group's discussion.

l’Reliability estimate after two dimensions were dropped

100

In general, initial estimates for all three constructs were rather low. Based on a
sample of 12 groups redundantly coded by the two raters who rated the construct, the
interrater reliability correlation for controversy ratings was .23. Similarly, the interrater
correlation of total scores generated by Raters 9 and 10 for process facilitation was .47
based on a second sample of 12 groups. Because of the large number of raters providing
judgments of intragroup conflict, two estimates were generated using four of the five
raters who made conflict ratings (the fifth rater only rated a few groups). The correlation
between Rater 2 and Rater 3 was very high (no, = .84) based on a sample of six groups
rated by both, but the correlation between Rater 4 and Rater 5 was extremely low (In =
.10) with a second, larger sample of 10 groups redtmdantly rated by these two individuals.

Thus, despite evidence during pilot testing that trained raters could produce
measures of intragroup conflict, controversy and process facilitation with adequate inter-
rater reliability, estimates for all three measm‘es obtained during the actual study were
tmacceptably low. In view of the low estimates of inter-rater reliability for intragroup
conflict, controversy and process facilitation, an effort was made to identify reliable
indices of each construct using a subset of the dimensions making up each construct. For I
all three constructs, using the samples involved in the initial estimate of reliability, new
scores were calculated for every possible combination of dimensions by summing the
unweighted dimension ratings involved in the composite. For example, with intragroup
conflict, new composite scores were formed by summing ratings on dimensions 1 and 2,
dimensions 1 and 3, and dimensions 2 and 3, as well as treating each of the dimensions as
a separate composite. After these new composite scores had been generated, interrater
reliability estimates were generated for each composite for each measure by correlating
the total scores of the new composites across the raters who provided the initial estimates.

Using this approach, it was possible to create an index of process facilitation using

only the clarifying and focusing/structuring dimensions with an acceptable level of

101
interrater reliability (rxx = .84). Of the two dimensions omitted from this new process
facilitation index, one was dropped because it never occurred (paraphrasing/
summarizing) while the second was removed due to a negative inter-rater correlation
among the two raters (integrating). It was not possible to form a more reliable index of
either controversy or conflict using a total score calculated from a reduced set of
dimension ratings.

Given the presence of the supplementary data for intragroup conflict, further effort
aimed at improving the measurement reliability of the judgmental variables was focused
on controversy. As no reliable sub-composite could be identified using a subset of the
component dimensions of controversy, the videotapes of group discussion were recoded
using a modified ratings format and a sample of each group's entire interaction. The
procedure devised involved single raters watching three five-minute samples of
interaction semi-randomly selected from each 75 minute tape (i.e., 20% of each group's
discussion phase). Four research assistants were involved in scoring the groups on this
revised measure, with one individual having provided original controversy ratings, two
individuals having provided intragroup conflict ratings, and one person having been
involved in rating process facilitation. The revised controversy measure is shown in
Appendix L.

The four research assistants available to recode controversy met with the primary
researcher for two training sessions at which the definitions of controversy and its
component dimensions were reviewed. Also during these sessions, eight group
videotapes were scored by all group members based on watching three five minute
segments of each group's discussion. After each group was observed, raters discussed
differences in their ratings and agreed on consensus ratings after collective discussion. At
the completion of the second session, the remaining unviewed groups were divided

among the four assistants and subsequent ratings were completed at the rater's homes.

102

An estimate of inter-rater reliability was generated by examining the matrix of
correlations based on the eight groups rated by three of the four assistants during training
(one assistant was not able to rate all groups during training). The median correlation
among the three pairs of raters was .56. In addition to this, after all groups had been rated
on the revised measure, the fourth assistant (left out of the above matrix) re-rated five
groups completed by one of the other three assistants chosen at random. The correlation
of total scores across the five groups for these two raters was .59. On the whole then, the
revised measure of observer ratings of controversy appears to have improved
measmement reliability to some extent but not to the point of traditionally-accepted levels
of measurement quality (i.e., .70 or better).

In addition to the observer ratings, data on group-level conflict were collected via
the post-experimental questionnaire as well except for one group which did not complete
this questionnaire by mistake. After two items were dropped, the eight-item
questionnaire measure of conflict produced an internal consistency reliability of Jo, = .78.

An AN OVA performed on the individual ratings of conflict was conducted with "group"
serving as the independent variable. The resulting test statistic [E (58, 164) = 3.10, p <
.00; MSW = 14.40, eta2 = .52] indicates that group membership had a significant effect
on individual scores and accounted for over half of the variance in individual scores.
This provides some support for the view that averaged individual conflict scores obtained
from the questionnaire measure can be meaningquy assigned to the group. As a result,
conflict scores were generated for each group by summing individual scores on the eight
remaining items and averaging the four individual scores within each group. However,
the weak correlation between the observer ratings of conflict and the averaged individual
questionnaire scores (5,, = .18) raises doubts as to whether these two measures are
tapping the same construct, although the poor reliability of the observer ratings is likely to

have severely attenuated this correlation. More will be said about this later.

103

In sum, despite pilot data which suggested the adequacy of the rating scales used
in this study, the reliability estimates for each of the three process measures obtained via
observer rating indicated room for considerable improvement with regard to the reliability
of the ratings measures. Although a reliable sub-composite of dimensions was
constructed for process facilitation, it was necessary to obtain a second measure of
controversy after revising the rating form and using only a sample of each group's
discussion. As it is possible to question the use of either measure of controversy or
conflict, multiple versions of each analysis were performed using each potential measure
when analyses involve either controversy or conflict (i.e., ratings and questionnaire data
for conflict; original and recoded ratings for controversy).

Manipulation Checks. The inquiry methods employed in this study were intended
to afiect a number of group processes during the experimental task. In particular, the two
conditions involving dialectical inquiry instructions were expected to produce more
controversy than consensus instructions, and the synthesis DI condition was expected to
yield more facilitative behavior than either of the other two inquiry methods. In actuality,
the omnibus F tests for the effects of inquiry method on both controversy [I_~‘ (2, 54) = .45,
p > .05] and process facilitation [E (2, 54) = .25, p > .05] failed to reach statistical
significance. However, the efl‘ect of inquiry method on controversy was marginally
significant when assessed using the recoded controversy scores (E (2, 52) = 2.78, p < .10;
_M_S_.m = 4.84; eta2 = .10). Planned comparisons among the inquiry methods (Mes =
3.92, Mm! = 3.83, M5151: 2.39) showed that the synthesis DI condition resulted in less
controversy than both the consensus condition (t (52) = 2.12, p < .05) and the traditional
DI condition (1(52) = 1.97, p = .05). The mean diflerence between the consensus and
traditional DI conditions was not significant.

An examination of the three questionnaire items pertaining to the quality with

which group members firlfilled their roles does little to explain the ineffectiveness of the

104
inquiry method interventions. These three items, shown in Appendix N, asked
individuals to rate on a 5-point Likert scale (1 = Strongly Disagree, 5 = Strongly Agree)
the extent to which they agreed with statements indicating: (1) they had effectively
carried out their personal assigned role, (2) they were conscious of that role during group
discussion, and (3) other group members had fulfilled their roles. Across all participants,
the means for these three items were 3.90, 2.31 and 4.22 (respectively), suggesting that
participants generally thought they had done what was asked of them and perceived their
fellow group members having done so as well. Further, participants generally reported
being conscious of their assigned roles during discussion (M = 2.31 for Item 2).
Therefore, in general, participants seemed to think they did a good job following
instructions.

With respect to the dialectical inquiry method conditions of particular interest
here, an examination of the first item corresponding to the respondent’s own role
performance has the potential to be more informative. However, consistent with the
previous data pertaining to all individuals, the two members assigned specific roles in the
TDI condition and all four members assigned roles in the SDI conditions generally
indicated that they thought they had done a good job. With regard to the TDI condition,
the mean response to the first item was 4.00 for the roles of both Vice President of Flight
Operations and Vice President of Industry Analysis. In the SDI condition, mean
responses across the four positions were 3.75 for Flight Operations, 4.20 for Industry
Analysis, 3.80 for Marketing and 4.05 for Finance. As a result, these data do not suggest
that the inquiry methods were implemented poorly.

Further, no group was removed from the analysis on the basis of the second
manipulation check, overall rater judgment. Raters were instructed to provide a
dichotomous, “yes-no” judgment as to whether groups should be removed for being
grossly incompetent or failing to take the task seriously. Given the already-insufficient

105
statistical power, raters were instructed to be conservative in recommending groups for
removal. Although no groups were subsequently recommended for removal, 3 number of
groups were found to be of bordrerline usefulness and might have been removed if data
had been available on a larger number of groups. However, on the whole, raters reported
that groups generally did know what they were doing and took their task seriously.

Overall, the inquiry method conditions appear to have been relatively inefl'ective
with regard to inducing controversy and stimulating facilitation in decision making
groups. Even further, the data suggest that the consensus-seeking approach resulted in
more controversy than the synthesis DI condition (but not more than the traditional DI
condition). The inquiry methods apparently did not impact process facilitation.

Nuisance Variables. In addition, a number of one-way AN OVAs were performed
as a check for the influence of several extraneous variables on the process and
performance variables of interest. The potential nuisance variables examined were
experimenter (five individuals), day of the week (every day except Saturday), time at
which the study was conducted (day, afternoon, or night), and type of participant group
(introductory psychology v. non-introductory psychology). Table 7 provides a summary
of the MANOVA conducted to test for the presence multivariate main effects stemming
from these four potential sources. As can be seen in the table, using Pillai's test statistic,
which tends to have the most power, no significant main effects were found for any of the
potential nuisance variables.

SW Table 8 displays the means, standard deviations, intercorrelations and
reliability estimates for measures of the variables used to test study hypotheses. As will
be recalled, there was a fair amount of attrition in groups resulting from one or more
pieces of missing data (see Table 5). For all correlations except those involving group
performance, sample size is 53 groups. Correlations in the last column of the table

involving group performance are based on 45 groups.

Table 7.

106

Multivariate Analysis of Potential Confound Variables.

 

 

 

 

 

 

 

 

 

 

 

Variable dfmum df mom Pillai's Approx. Sig.
Value F Ratio
WEEKDAY 5 60 2.15 1.30 .19
TIMEOFDAY 2 18 1.11 1.59 .18
EXPERIMENTER 4 44 1.21 .68 .86
SUBJECT LEVEL 1 8 .56 1.44 .31

 

Note: Multivariate statistic reported is Pillai's trace.

 

Table 8.

Means, Standard Deviations, Intercorrelations and Reliabilities for Measured Variables.

107

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

ABIL KNOW CONTl CONT2 FACL CONFl CONFZ
Mean 95.40 9.77 3.91 3.46 3.54 2.17 15.03
SD 11.18 3.18 1.13 2.26 1.31 1.53 3.42
R,“ — - .23 .55 .85 .10/.84 .78
ABIL .12 —.02 -.12 -.02 -.25 .15
KNOW .11 .06 .24 .06 -.08
CONTl .30 .19 .19 .06
CONT2 .12 .30 .34
FACL .14 .11
CONFl .18
CONFZ
INFO
#MEN
PERF

 

Note: N = 53 for all correlations except those involving PERF (N = 45)

Maugham:

ABIL = Sum of Cognitive Ability scores in group (Wonderlic Personnel Test)

KNOW = Sum of individual Role Knowledge Scores in group

CONT] = Observer rating of group-level Controversy

CONT2 = Recoded observer rating of group-level Controversy
FACL = Observer rating of group-level Facilitation
CONFl = Observer rating of group-level Conflict

CONFZ = Averaged group member questionnaire responses for Conflict

INFO = Sum of unique pieces of information shared by all members

#Men = Number of males in the group
PERF = Group Performance (net value)

 

Table 8 (cont’d.).

Means, Standard Deviations, Intercorrelations and Reliabilities for Measured Variables

 

 

 

 

108

 

 

 

 

 

 

 

 

 

 

 

INFO #MEN PERF
Mean 35.62 1.53 121.12
SD 13.66 1.08 50.00
Ru .... _ ..
ABIL .23 - 03 34
KNOW .l l .05 .34
CONT] .13 .22 .03
CONT2 .01 .08 -.02
FACL .19 -.01 .37
CONFI .04 -.01 -.1 l
CONF2 -.07 -.01 .06
INFO .17 .05
#MEN .09
PERF

 

 

 

 

 

 

Note: N = 53 for all correlations except those involving PERF (N = 45)

Abbreviations:

ABIL = Sum of Cognitive Ability scores in group (Wonderlic Personnel Test)
KNOW = Sum of individual Role Knowledge Scores in group

CONT] = Observer rating of group-level Controversy

CONT2 = Recoded observer rating of group-level Controversy

FACL = Observer rating of group-level Facilitation

CONFl = Observer rating of group-level Conflict

CONF2 = Averaged group member questionnaire responses for Conflict
INFO = Sum of unique pieces of information shared by all members
#Men = Number of males in the group

PERF = Group Performance (net value)

109

Process Hypotheses

Information Sharing x Intragoup Conflict --> Group Performance. Hypothesis 1,
predicting an interaction effect between unique information sharing and intragroup
conflict on group performance, was tested with a hierarchical moderated multiple
regression analysis. In order to examine the incremental contribution of process over and
above knowledge and ability in predicting group performance, the average cognitive
ability and task knowledge scores for each group were first entered into the equation at
Step 1. Following this, unique information sharing and intragroup conflict were entered
at Step 2 followed by the interaction term (Intragroup Conflict x Information Sharing) at
Step 3. Support for Hypothesis 1 would be obtained by noting a significant increment in
r2 at Step 3 when the product term is added to the equation after its additive components
have already been entered at Step 2. Because of the two sources of conflict data
available, this analysis was conducted twice, once with the ratings data and once with the
questionnaire data. Table 9 summarizes the results of the regression analyses associated

with Hypothesis 1.

110

Table 9.
Hierarchical Moderated Regression Results for Hypothesis 1.

Using observer ratings of conflict:

 

 

 

Step Vamblets) Added r2 change A

1. Ave. Cognitive Ability .24 .00
Ave. Role Knowledge

2. Information Sharing .00 (.02) .96 (.60)
Intragroup Conflict

3. Information Sharing x .04 (.05) .14 (.13)
Intragroup Conflict

Using averaged individual perceptions of conflict:

 

 

 

 

Stet; Variable! 8) Added r2 change 9

1. Ave. Cognitive Ability .22 .00
Ave. Role Knowledge

2. Information Sharing .01 (.01) .85 (.81)
Intragroup Conflict

3. Information Sharing x .00 (.00) .95 (.79)
Intragroup Conflict

Notes:

N = 49 groups

Values in parentheses apply when control variables are not entered on Step 1

1 1 1

As can be seen in the table, group cognitive ability and task knowledge account
for a good share of the variance in group performance while information sharing and
conflict do not add anything to the predictive accuracy of the equation. Using observer
ratings of group conflict, there is a .04 increment in 1'2 when the interaction term is added
to the equation at Step 3 (t (43) = -l .50, p > .05). When the questionnaire data regarding
conflict are used instead, there is clearly no interaction effect (r2 change = .00). When
these two analyses are conducted without entering the cognitive ability and task
knowledge, the change in r'2 accompanying the interaction term increases to .05 for the
observer ratings of conflict (1 (45) = -1.53, p > .05) but remains at .00 for the
questionnaire data. Although more will be said about this interaction in the Discussion,
there is no support for Hypothesis 1 predicting that a primary determinant of group
performance is the interaction between information sharing and conflict.

Controversy --> Information Sharing, Intragroup Conflict. Hypotheses 2 and 3
predicted that controversy would be positively related to both information sharing
(Hypothesis 2) and intragroup conflict (Hypothesis 3). Using the original ratings of
controversy (CONTl in Table 8), the group-level correlations with unique information
sharing (r (53) = .13, p > .05) and intragroup conflict (CONFl) as measured by observer
rating (r (53) = .19, p > .05) were in the expected direction but not large enough to reach
statistical significance. When the questionnaire data were used as the measure of conflict
(CONF2), the corresponding correlation between controversy and conflict was also non-
significant (r (53) = .06, p > .05).

On the other hand, the correlations between controversy and both unique
information sharing and controversy and intragroup conflict changed somewhat when the
recoded controversy scores (CONT2) were used instead of the original observer ratings
(CONTl ). With the more reliable CONT2 measure, the observed correlation between
controversy and information sharing decreased somewhat (r (53) = .01, p > .05), but the

1 12
correlation between controversy and observer ratings of conflict increased to the point of
statistical significance (r (5 3) = .30, p < .05). Similarly, the correlation between the
recoded controversy scores and questionnaire-based conflict measure also attains
statistical significance (r (53) = .34, p < .05).

Overall, there was support for Hypothesis 2 regarding the positive relationship
between controversy on group conflict when the more reliable recoded controversy scores
were used, but there was no support for Hypothesis 3 predicting that higher levels of
controversy within groups are associated with higher levels of unique information
sharing.

Process Famation -> Information Sharing, IntragQup Conflict. Hypotheses 4
and 5 predicted a negative relationship between process facilitation and intragroup
conflict (Hypothesis 4) and a positive relationship between process facilitation and
information sharing (Hypothesis 5). As with controversy, process facilitation (F ACL in
Table 8) was not significantly correlated with unique information sharing (INFO), but the
observed correlation was in the expected direction (a (53) = .19, p > .05) . On the other
hand, the observed correlation between process facilitation and intragroup conflict was in
the opposite direction from that predicted (although non-significant) for both the observer
ratings measure of conflict (1(53) = .14, p > .05) and the aggregated group member
questionnaire responses (r (53) = .11, p > .05).

An examination of the correlations between the two component dimensions of
facilitation and the two overall measures of conflict revealed that the two dimensions
both had zero or positive relationships with conflict, but each was more strongly related
to a difi‘erent measure of intragroup conflict. Average group scores on the questionnaire-
based index of conflict (CONF2) were more strongly correlated with ratings on the
"clarifying" dimension of process facilitation (r (56) = .14, p > .05) than with ratings on

the "focusing/strumming" dimension (; (57) = .04, p > .05), while observer ratings of

1 13
conflict (CONFl) were more strongly correlated with the "focusing/structuring"
dimension (g (57) = .20, p > .05) than with the "clarifying" dimension (; (56) = .00, p >
.05). Although none of these correlations is significant and the difference in pattern may
simply be the result of sampling error, it may be that the two measures of conflict are
tapping somewhat different constructs.

In sum, there is not much support for the notion that process facilitation impacted
the degree of conflict and information sharing within these groups. Although the
obtained correlation between facilitation and information sharing was positive, it was not
large enough to be statistically significant. On the other hand, the positive correlation
between facilitation and conflict was not predicted but also not large enough to reach
significance. An examination of the relationship between the two component dimensions
of process facilitation (clarifying and focusing/structuring) and the two measures of
conflict (CONFl and CONF2) did not find either dimension of process facilitation to be
negatively related to conflict, but did uncover a somewhat different pattern of
relationships between the two dimensions of process facilitation and the two measures of
conflict, suggesting that observer ratings of conflict and the aggregated group member
perceptions of conflict did not measure the same thing.

Ingg'g Method Hymtheses

Hypotheses 6 and 7 predicted a relationship between inquiry method (CS, TDI,
SDI) and two process variables (information sharing and intragroup conflict), while
Hypothesis 8 concerned the impact of inquiry method on group performance. These three
hypotheses were tested using one-way AN OVAs with inquiry method as the independent
variable and information sharing, conflict and group performance (respectively) as the
dependent variables with group as the level of analysis. Table 10 provides summaries of
these three AN OVAs and Table. 11 reports the means and standard deviations for the
three inquiry method conditions across the three dependent variables.

1 14

As shown in Table 10, F-tests for the effects of inquiry method on all three
dependent variables were not significant. For information sharing [E (2, 54) = .40, p >
.05] and group performance [E (2, 48) = .84, p > .05], the value of the F statistic was
below 1.00. The F -test for the AN OVA on conflict was marginally significant when
observer ratings (CONFl) were used [E (2, 54) = 2.40, p < .10], but not with the
questionnaire-based conflict scores (CONF2), g (2, 55) = .39, p > .05. Based on the
marginally significant F-test for the conflict scores, the planned comparisons implied in
Hypothesis 7 were carried out. Using observer ratings of intragroup conflict (CONFl), as
predicted, the mean level of conflict in the synthesis DI condition (M = 1.50) was lower
than mean for the traditional DI condition (M = 2.58) , t (54) = 2.71, p < .05). The other
two planned comparisons were not significant.

As a result, there is no support for Hypotheses 6 or 8 regarding the superiority of
synthesis DI over traditional DI and consensus approaches with regard to information
sharing or group planning quality. However, there is partial support for Hypothesis 7 and
the prediction that synthesis version of D1 would result in less conflict than traditional DI
when observer rating of intragroup conflict are used as the dependent variable (as

opposed to aggregated group member perceptions of intragroup conflict).

Table 10.

ANOVA Summaries, Hypotheses 6-8.

115

 

 

 

 

 

 

 

 

 

 

 

 

 

DV drum, dam, F ratio Sig. Eta2
— _—_
Group 2 48 .84 .44 .03
Performance
Information 2 54 .40 .67 .0]
Sharing
Conflict 54 2.40 .10 .08
Conflict 55 .39 .68 .01
(questionnaire)
Table 11.

Cell Means and Standard Deviations, Hypotheses 6-8.

 

 

 

 

 

 

 

 

 

 

 

 

I DV Consensus Traditional DI Synthesis DI I
Group Performance 131.32 114.26 L 109.54
(62.67) (42.64) (46.14)
Information Sharing 33.26 37.21 34.89
(14.27) (10.88) (15.39)
Conflict (ratings) 2.08 2.58 1.50
(1.22) (1.93) (1.31)
Conflict 14.89 15.58 14.62
(Que-“mm!” (3.75) (3.37) (3.23)
m:

Cell sizes for Conflict and Information Sharing: 19 groups
Cell sizes for Group Performance: 16 (CS), 17 (TDI), 18 (SDI)

 

 

116

Surmnag
Although the inquiry method interventions appear to have been weak, there is

some support for several study hypotheses. The synthesis DI condition does appear to
have resulted in significantly less conflict than the traditional DI condition as predicted,
but did not yield more facilitation than either of the other two conditions. Further, the
consensus condition resulted in more controversy than the synthesis DI condition, but not
more than the traditional DI condition. With regard to the efieas of process variables on
other process variables, generally weak and nonsignificant effects were found that tended
to be in the predicted direction. In particular, the recoded controversy scores (CONT2)
were found to be significantly and positively correlated with the both measures of group
conflict. Finally, the predicted interaction of information sharing and conflict was not
significant but did produce a .04 change in r2 when entered into the regression involving
group performance after controlling for group-level cognitive ability and task knowledge.
Figure 4 displays estimates of the path coefficients (beta weights) for the
combined process model pictured in Figure 2. These estimates were obtained from three
separate single-step regressions, one for each of the dependent variables in the model
(i.e., conflict, information sharing and group performance). Because of the improved
reliability, the recoded measure of controversy was used in this analysis. As can be seen
in Figure 4, the only statistically significant path is from controversy (recoded) to conflict
(observer ratings), but most coeficients are in the direction predicted by the model. Of
particular note, the path from controversy to conflict does not change substantially
regardless of the conflict measure used (.32 for ratings, .33 for questionnaire data).
Again, the interaction effect predicted in Hypothesis 1 is strong enough to warrant
consideration, but is not significant at the .05 level. The prospects for this model will be

a central theme in the next section.

117

E52 388m 2: com 3383mm 5mm .v 2&3

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

mo.

 

samba/8:80

 

C.

 

 

oocmgofiom wccgm
@386 A > :BEESE:
#0. H NH afio .
mm
85:00 meccwebfi A
S.

 

sessions
388m

 

 

1 18
Exploratory Analyses

Given the findings by Stasser suggesting that biased information sampling is an
extreme problem in decision making groups, it is interesting to consider the extent to
which groups identified the unique informational resources of their members in this
complex, ill-structured setting. Unfortunately, given that a different task was used, it was
not possible in this study to compare the degree of bias present in the information
sampled at the group level with previous work.

At a purely descriptive level, across all conditions, groups shared on average
35.62 unique information cues out of the 510 1mique cues available to members - only
7%. On the other hand, during discussion groups shared an average of 39.79 cues known
beforehand by all members out of a total of 190 common information cues - 21%.
Although the absolute values of these percentages are rather meaningless given the ease
with which they can easily be manipulated by task demands, the relative difference in
rates is striking. Basically, on average, groups were three times more likely to mention an
information cue known to all members before discussion than an information cue known
only to one member (or in a few cases, two members). This is remarkably similar to the
2.5:] ratio observed in earlier work by Stasser et a1. (1989).

In light of this, a follow-up analysis was conducted to examine whether the
inquiry methods used in this study had any efl‘ect on the relative amounts of sharing for
unique and common information cues. To this end, a new variable was derived for each
group by dividing the number common information cues mentioned by the number of
unique information cues shared. This ratio, which represents the number of common
information cues mentioned for every unique piece of information shared, was then used
as the dependent variable in an AN OVA with inquiry method serving as the independent
variable. The resulting F ratio was marginally significant [F (2, 53) = 2.52, p < .10;

MSW = .28, eta2 = .09], but no two conditions were significantly different using Tukey's

l 19
honestly-significant difference test for post-hoe comparisons (M = .96, 1.15, 1.35 for CS,
TDI and SDI conditions, respectively).

A final exploratory analysis was conducted to examine the degree to which groups
were able to uncover the "hidden" nature of the optimal strategy. The task was
constructed so that, although several difl‘erent strategies could be employed to increase
profit, the one which would yield the most profit (according to expected values for route
revenue using the most likely level of competition) was the expansion strategy. As noted
earlier, the only group member that had any information relevant to expansion was the
Vice-President of Industry Analysis. One possible indicator of the extent to which groups
followed an expansion strategy is in terms of the number of new routes added to flight
operations. To examine the possibility that the DI methods may have led groups to adopt
the expansion strategy more than the CS method, an AN OVA was conducted with the
number of new routes added to the flight plan as the dependent variable and inquiry
method as the independent variable. The resulting F test was non-significant [E (2, 57) =
1.40, p > .05], suggesting that the dialectical inquiry condition was not more successful
than the consensus condition in leading groups to "tmcover" the optimal strategy. I will
have more to say about the role of biased information sampling in complex, ill-structured

decision making tasks in the Discussion.

DISCUSSION

Study Contributiona

The primary purpose of this study was to examine the determinants of group
decision making in a complex, ill-structured decision making task with particular
attention to the incremental contribution of process-related factors over and above group
input variables such as cognitive ability and task knowledge. Beyond this, the effects of
several possible group decision making structures (i.e., inquiry methods) were also
assessed with regard to their expected impact on group process and performance. A third
purpose of this study was to examine the phenomenon of biased information sampling in
a complex, ill-structured environment where group members represented "experts" from

difl‘erent areas of an organization.

Overall, there was little support for the proposed model. More precisely, only two
hypotheses involved statistically significant effects «the relationship between controversy
and intragroup conflict (Hypothesis 3), and the effect of inquiry method on intragroup
conflict (Hypothesis 7). However, given the low statistical power in this study, it is
important to note that the obtained pattern of effects was generally consistent with the
predictions of the model. The results of this study, qualified though they must be, suggest
that group process variables may have some incremental validity over and above the
powerful input factors of ability and task knowledge. Of particular interest was the
interaction of information sharing and intragroup conflict. This interaction continued to
account for 4% of the variance in group performance even after controlling for group-

level cognitive ability and task knowledge. In addition, the measure of process

120

12 1
facilitation unexpectedly had the strongest relationship with group performance ~-
stronger even than cognitive ability and task knowledge. These findings provide some
support for the notion that process variables such as intragroup conflict, information
sharing and process facilitation may explain some of the variance in group performance
that group-level cognitive ability and task knowledge cannot.

In the remainder of this section, I discuss a number of issues related to evaluating
and refining the model proposed in the introduction and the use of structural
manipulations (i.e., "inquiry methods") to improve group performance. I also comment
on the issue of biased information sampling in complex, ill-structured decision making
environments and conclude by identifying a number of areas that need further research
attention as a model of group decision making in ill-structured contexts is progressively
identified.

Study Limitations

Given theoretical basis for the hypotheses in this study, an important issue to
address is the general lack of support for the model. In this section, I first discuss several
measurement issues that constrain conclusions about the model, then discuss the overall
model with these issues in mind.

Boundary Conditions. As noted earlier, due to the nature of the task employed in
this study, the current findings are limited to groups in situations involving non-routine
decisions, face-to-face interaction, and large amounts of task-relevant information
distributed across group members. Although some organizational decision making
groups certainly operate in these conditions, others clearly do not. Given these task
characteristics, the results of this study are most applicable to temporary, ad-hoc,
heterogeneous decision making groups dealing with complex, ill-structured, strategic
problems. In addition, the use of undergraduate students in the present study further
limits generalization to real-world strategic decision making groups. Undergraduate

122
students in psychology differ in many non-trivial ways from experienced business
managers and executives. For example, undergraduates are likely to have had less
experience making decisions in groups and less opportunity to have developed a sense of
their own competence in such situations. Difl‘erences such as these do not necessarily
mean that the findings in this study would not be found in other samples. However, they
do require that caution be used when generalizing results to other populations. Clearly,
the efl‘ects found in this study need to be replicated in the field with actual managers and
existing work groups.

Mement Iasues. There are several measurement issues which necessitate
cautious interpretation of study results: (1) Measurement reliability, (2) Range
Restriction, (3) Construct validity, and (4) Low statistical power.

To begin with, poor measurement reliability proved to be a troubling and
inn'actable problem for all three judgmental measures of group process used in this study
(i.e., process facilitation, controversy and intragroup conflict). Despite piloting and
training, initial reliability estimates for observer ratings of conflict, controversy and
facilitation were all below .50. Subsequent efforts to improve the reliability of these
variables were only partially successful. The efl‘ect of poor measurement reliability is to
attenuate the observed correlation between two measures, with the decrement in the
magnitude of the observed correlation multiplicatively worsened when both measures
have low reliability. As a result, correlations involving the two observer ratings of
conflict and controversy are likely to be underestimated.

Although it is certainly possible to correct observed correlations for attenuation
due to measurement error, the appropriateness of this procedure is heavily dependent on
the accuracy of the reliability estimates. A number of concerns call into question the
accuracy of the reliability estimates generated in this study, including the small samples

sizes available for the interrater reliability correlations and the different estimates for

123
observer ratings of conflict based on two independent pairs of raters. Ideally, in situations
involving interrater reliability, a matrix of inter-rater reliability coefficients would be
available involving all possible pairs of raters for each measure, and there would be little
variance across the set of pairwise estimates such that the mean (or perhaps median)
correlation would provide a stable estimate of the overall reliability for the measure. For
practical reasons, it was not possible to generate such a matrix for this study except for
the recoded controversy measures (CONT2), and therefore all reliability estimates (except
one) are based on only one of several possible pairs of raters for each measure. In the
case of the exception, observer ratings of intragroup conflict, the two estimates available
differ considerably (. 10 versus .84). All things considered, correcting the observed
correlations in this study for measurement error may very well be misleading.

One factor which may have coincidentally contributed to the low levels of
reliability for all three measures based on observer ratings is range restriction on their
corresponding constructs. An examination of Table 8 reveals that the means and standard
deviations for the three process measures (CONTl, CONFl and FACL) are quite small in
magnitude compared to their respective maximum possible scores of 9.0, 9.0 and 12.0.
The notion of range restriction is consistent with the anecdotal impressions of several
raters, who noted long periods of silence during some group's discussion time and a
general low-keyed tone to many group discussions. After observing the videotape of a
particularly quiet group, one rater was moved to inquire "Did they know they could talk?"

Although high levels of process facilitation, intragroup conflict and controversy did
occm' within some groups during the study, across all groups the distribution of scores on
these three measures was positively skewed. This problem was anticipated and steps
were taken to address this issue in the design of the task (e.g., divergent information, role
guidelines, monetary rewards, etc.), but it appears that what was done was not sumcient

to overcome the strong cultural norms against talking to strangers, appearing disagreeable

1 24
and/or looking stupid. Although incentives and experimental procedures can offset this
to some degree, the root of the problem may lie more with the nature of the people
involved (i.e., undergraduate psychology students) than the design of the task.

One potential design mechanism that might ofiset the motivation loss inherent in
group norms inhibiting disagreement would be the existence of incompatible individual
sub-goals in addition to an overall group goal. With respect to the task used in this study,

along with being told to create a plan yielding maximum profit for the organization, each
Vice President might be given one or more sub-goals related to incorporating certain
features into the final group plan considered to be important by the department he or she
represents. For example, the Vice President of Flight Operations might be told to see that
annual fuel costs are held below a certain amount, or the Vice President of Finance might
be assigned the sub-goal of reducing existing personnel levels by 10%. With
corresponding incentives for their achievement, individual sub-goals arrayed in a trade-
ofl‘ fashion could act as conflict “lightning rods” by providing tangible issues arormd
which amorphous disagreement could coalesce. Given the seemingly ubiquitous
presence of non-aligned goal hierarchies in real-life strategic group decision making,
directly incorporating conflicted sub-goals in the design of a research task would increase
the realism and generalizability of study findings. As a result, future research should
strongly consider employing individual sub-goals along with an overall group goal.

Although the use of multiple measures of the various process constructs was able
to alleviate some of the problem associated with low reliability, using these alternative
measures may have also created a construct validity problem in that the revised measures
may not have tapped the same construct domain as the original measures. This is most
pronounced in the case of facilitation, where the final scale used consisted of only two of
the four dimensions defined as part of process facilitation. Although respectable

reliability was attained using two of the four dimensions, the modified measure of process

125
facilitation is now somewhat deficient compared to the original conceptualization of the
construct. In spite of this deficiency, it is interesting to note that the revised process
facilitation measure has a relatively strong relationship with group performance. This
relationship might be even stronger with a reliable measure of the broader construct.

In the case of conflict, the correlation between the two measures of the construct,
although most likely attenuated by the low reliability of the ratings data, is only .18.
Further, an examination of the pattern of correlations for the respective conflict measures
and other measures reveals marked differences. In several cases, the correlations between
another measure and the two measures of conflict (CONFl and CONF2) are in opposite
directions (i.e., cognitive ability, task knowledge, information sharing and group
performance). One notable exception to this pattern is the relatively strong and positive
correlation both conflict measures have with the recoded controversy measure (CONT2).

Still, the low correlation between the two conflict measures and the different pattern of
external correlations suggests that the two measures of conflict are not tapping the same
construct. Looking at the correlations with other measures, the pattern for the observer
ratings of conflict makes the most theoretical sense. Given the post-hoe, self-interested
nature of having group members rate the conflict in their own groups, it is tempting to
treat the observer ratings of conflict as the better (if less reliable) measure for theoretical
reasons. Unfortunately, the issue cannot be definitively resolved in this study and
remains a complicating factor in attempting to identify the relationship between
intragroup conflict and other group process constructs.

Conversely, the two sets of controversy ratings (CONTl and CONT2) show more
convergence than the two conflict measures. The convergent validity correlation between
the two measures of controversy is relatively strong (5,, = .30) given the low reliability of
the initial (CONTl) ratings (an i: .23), and the pattern of correlations with other measures

is generally in the same direction. On the other hand, the original ratings of controversy

126
(CONTl) are more strongly related to facilitation and information sharing, while the
recoded controversy measure (CONT2) is strongly related to the questionnaire-based
measure of group conflict and the original measure is not. Although it seems reasonable
to use the recoded controversy scores given the improved reliability and the generally-
similar pattern of outside correlations, it is likely that the two measures are not tapping
the same construct domain.

A final measurement-related problem for this study which is common to many
studies involving groups is the low statistical power associated with the relatively small
sample size. Given that all path coefficients but one were in the predicted direction, it is
possible that more linkages would have been statistically significant had there been
greater statistical power for the analyses. The probability of obtaining significant results
with two-tailed tests and alpha = .05 is only 11% when rho = .10 and stilljust 30% rho =
.20 (Cohen, 1988). Ifone is searching for relatively small effects associated with
complex, overdetermined phenomena (as likely in group decision making), it would take
approximately 200 groups to yield an 80% chance of achieving significance when rho =
.20 and almost 800 groups when rho = .10! Given the logistical difficulties associated
with collecting that much data on groups, there often may be no alternative to conducting
studies that are seriously underpowered in some respects.

In sum, there are a number of measurement issues which cloud interpretation of
the results obtained in this study. However, the low reliability of a couple of variables,
range restriction and low statistical power make this study a conservative "test" of the
model. In some respects, given these obstacles, it is noteworthy that any substantial
efl'ects were found at all. It is to these effects which we now turn.

The Process Model

 

Overview. In general, the process model hypothesized in Figure 2 was not
supported in this study. The only hypothesized linkage found to be statistically

127
significant was between controversy and intragroup conflict. The one significant
relationship that was not predicted but which is relevant to the model is that between
process facilitation and group performance. On the other hand, the pattern of obtained
relationships was generally as predicted in the hypotheses. This suggests that, consistent
with the previous discussion, a primary reason for the lack of support for the model is low
statistical power. In particular, the hypothesized interaction of information sharing and
conflict in affecting group performance was not significant in spite of accounting for 5%
of the variance in profit earned by the groups in the simulation. Given that the interaction
was hypothesized and the change in r2 was large enough to be of substantive interest, we
now turn to a closer examination of this interaction.

Interaction of Information Sharing and Conflict. Given the low power associated
with this analysis and the predicted interaction, it is useful to plot the interaction between
unique information sharing and intragroup conflict on group performance in order to
determine if the interaction could be interpreted in a fashion consistent with the
prediction in Hypothesis 1. Using the regression equation without the control variables
entered on Step 1, the interaction is plotted in Figure 5. To generate the lines shown in
Figure 5, specific values of information sharing and conflict were selected and inserted
into the regression equation. The graph shows that, when intragroup conflict is low (i.e.,
-1 SD on intragroup conflict), high levels of information sharing were associated with
higher levels of profit in the simulation. However, when there is a high level of conflict
in the group (+1 SD on intragroup conflict), a high level of information sharing is
associated with lower profit. It is also important to note that the variance accounted for
by the product term is essentially independent of the variance accounted for by mean
individual cognitive ability and task knowledge. Thus, although the effect did not reach

traditional levels of statistical significance, the strong effect size and predicted nature of

128

55385 85.50 anew—«:5 x wstmcm cosmetomfi .m enema

05.65 5sz.22
:9: 2,3

 

 

 

29:80 :9: I
8580 33 l

 

 

 

eeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee

IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII

IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII

IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII

 

 

 

A3053 5 cto>> #mzv 856.0th 9.65

cm
ow
om
om
oo P
ON F
0v _.
om P

129
the interaction suggest that future research should examine this interaction again using a
design with more statistical power.

Revised process model. In general, this study did not provide much support for
the proposed model of group decision making in ill-structured task environments.
However, as noted, the existence of low reliability of two measures in the model, range
restriction and low statistical power provided very conservative conditions under which to
test the model. As a result, it is not clear that this model should be dismissed just yet. At
the same time, it is useful to identify the factors which were most useful in explaining
group performance in this complex, uncertain task. Measurement issues not
withstanding, Figure 6 presents a modified model of group decision making in
ill-structured tasks.

Overall, the revised model highlights the dual importance of "input" and "process"
factors in determining group performance on ill-structured tasks. Whereas the a priori
model focused largely on the role of process-related factors and treated cognitive ability
and task knowledge as control variables, this model directly incorporates them on the
basis of their strong and relatively independent efi'ects on group performance. What may
have happened in this study, as it may in many similar real-life situations, is that input
factors such as ability and task-related knowledge manifest themselves independent of
group process through the identification of high-quality recommendations and
alternatives on the part of individual group members. Clearly, individual
recommendations and suggestions can incorporate unique, task-related knowledge
without explicitly referring to that knowledge (i.e., sharing unique information). Thus, to
some extent, the individual inputs of group members (e. g., cognitive ability and
specialized knowledge) may impact group performance without involving group
interactive processes. However, beyond a certain point, it is probably the case that

increments to group performance are a function of group process (i.e., discussion and

130

mac—02 003009 0:80 .00 3002 0830M < 0 80w?—

 

 

00§EB§ 00.5

 

 

 

 

 

 

 

 

 

0w0230§ 0331
0000800005 \

 

025.4 essence

 

0:000

 

 

 

u e warmsm .0000

 

 

 

 

 

 

 

005000 00000005

 

 

 

tombs/00:00

 

 

 

 

13 1
interaction). At this point, process-related factors such as facilitation, information sharing
and conflict become important. The relative weights for the input factors and process
factors may then depend on the exact nature of the particular task.
.Dlrlectical Inquiry in Group DecisLmr MM

In general, the DI methods used in this study do not appear to have had much
impact on group process. Although the DI methods were explicitly intended to impact
facilitation and controversy, they do not appear to have done so (although synthesis DI
did result in less group conflict than traditional D1). The self-report data gathered via
questionnaire suggested that group members in the role of plan presenters and debaters
thought they were doing their job. Why did the two DI methods not increase controversy
and why did the synthesis DI not increase facilitation? Previous research has formd that
controversy-inducing methods such as DA and DI result in higher levels of group
performance than the consensus-seeking method. Why was this not found in the current
study? The explanation may reside in the particular manner in which DI was
implemented in this study.

First, it should be recalled that the task was explicitly designed so that group
members would have to deal with divergent information, incompatible alternatives and
necessary trade-offs. In other words, there was a great deal of latent controversy built-in
to the task and the two dialectical inquiry methods were designed to bring this
controversy to the surface. The two members asked to present and critique plans were

’ “ration pointing to the need for radically difi‘erent changes to the existing plan.
members were asked to question the other presenter as to the

ations and were not supposed to stop until achieving a good

person's position. Detailed instructions were provided to the

edure they were to use to implement the dialectical methods,

l 32
and DI group members received specific and comprehensive instructions regarding how
they were to fulfill their respective role assignments.

Within this context, there are several factors that seem likely to have played a role
in the failure of the DI methods to produce their expected consequences. First, as alluded
to previously, participants on the whole did not know one another before participating in
the study and appear to have been reluctant to engage in behaviors that could have been
considered confrontational. Second, with regard to stimulating controversy, the particular
manner in which both DI techniques were implemented probably lessened the
controversial impact. Specifically, the simultaneous creation of two independent plans
rather than the iterative creation of a plan and counterplan did not allow
presenters/critiquers time to familiarize themselves with the opposing plan (essentially
requiring all arguments to be generated on-line). Third, despite efforts to insure that
planners created incompatible plans, the simultaneous ("blind") generation of competing
plans may have resulted in perspectives that were not always antithetical. Previous
research suggesting DI improved group performance relative to CS has always used plans
that were created in advance and presented intact (assigned) to group members who then
advocated them in the dialectical process. Combined, these factors suggest thatthe DI
methods did not yield more controversy and higher group performance in this study
because too much of the dialectical process was assumed to occur rather than forced to
occur.

The failure of the synthesis DI condition to produce more process facilitation is
more difficult to explain, as it was not predicated on antithetical plans or confrontational
interpersonal behavior. Here, social factors appear to be most relevant in explaining the
lack of predicted efi‘ect. Enacting the synthesis role required the ability to comprehend
and integrate what others were saying, and necessitated the willingness and ability to

1 3 3
provide leadership. This may have been beyond the motivation or capabilities of some
participants.

In summary, a variety of factors may have combined to reduce the expected
effectiveness of the DI inquiry methods in producing controversy, facilitative behavior
and subsequently group performance. Although generating and assigning antithetical
plans to groups in the DI conditions might have improved the strength of these
manipulations, it would have eliminated the opportunity to observe the degree to which
DI methods allowed groups to "uncover" the hidden profile of the correct strategy (i.e.,
expansion). Did groups using DI uncover the better (i.e., expansion) strategy more often
than the CS groups? It is to this question that I now turn.

Biased Information Sampling and Hidden Profiles

Despite the diflemnce in tasks, it is interesting to compare the relative rates of
information sharing obtained by Stasser et al. (1989) with those found in this study. In
the earlier study, groups were approximately 2.5 times more likely to mention an
information cue known to all members than an information cue known only to one. In the
current study, groups were three times more likely to do so. As a result, initial indications
are that the bias towards shared information is fairly robust across task type.

Further, in keeping with recent research by Stasser (1992) and Stewart and Stasser
(1995) which formd that "advocacy" and "expert role assignment" manipulations
(respectively) were not particularly effective at reducing the bias in favor of common
information, the dialectical inquiry methods used in this study did not result in more
unique information sharing or a lower ratio of common/unique sharing during group
discussion. Indeed, consistent with Stasser's (1992) results with the DISCUSS
simulation, this study found that the bias in favor of common information may be
worsened by conditions that require one or more group members to "advoca " a position

(i.e., "plan" v. "counterplan"). As a result, it is still unclear how DA and DI have their

1 34
beneficial effects on group performance. One possibility is that these methods cause
more alternatives to be generated and considered by the group without inducing group
members to share relevant supporting information. Future research might further explore
the mechanism by which DA and DI impact group performance by measuring the number
and quality of individual recommendations, the number of alternatives considered by
groups, and the relative amounts of common and unique information shared during group
discussion.
Futulre Direction_s

There are a number of issues raised by this study which warrant attention in future
research. First, future research should consider using ill-structured, moderate-fidelity
business simulations that can be conducted in a relatively short period of time. In the
past, research on group decision making has tended to use tasks that are relatively
simplistic and unengaging (e.g., the Moon Survival task or simple case studies) or tasks
that are extremely complex and intended primarily as teaching tools (e.g., semester-long
management simulations). As this study demonstrated, there is a middle grormd between
the two extremes that can present participants with a challenging, self-contained task
environment that is probabilistic, complex and moderately realistic.

Given the low statistical power and the measurement issues present in this study,
further research should continue to address the adequacy of the model of group decision
making proposed in this study. In particular, with improved measurement reliability and
greater statistical power, it should prove interesting to compare the original model from
Figure 2 with the revised model presented in Figure 6. With a larger sample size and
better measurement reliability, it would be possible to directly compare these two models
with structural equations modeling techniques.

Future research might alSo better address several issues related to levels of

analysis. Roberts, Hulin and Rousseau (1978), among others, have identified the need to

135
consider and examine multiple levels of analysis in order to understand the behavior of
individuals in organizations. When organizations are viewed as being composed of
hierarchical systems operating at multiple “levels” of social complexity, it becomes both
possible and necessary to understand organizational performance as a function of
processes occurring at and across different levels of the organization through the use of
composition, cross-level and multi-level theories (Rousseau, 1985). Two levels of
analysis particularly relevant to tmderstanding the performance of decision making

groups are the group and the individual. However, when data are measured at one level

and analyzed at another, it becomes necessary to aggregate/disaggregate measuremen --
a procedure that can artifactually create or alter functional relationships between focal
constructs and other constructs (James, 1982).

Because the focal level of interest in this study was the group, constructs were
(for the most part) conceptualized, measured and analyzed at the group level in an efl‘ort
to avoid the potential problems and ambiguities associated with aggregation. One
exception to this statement concerns the composition variables used as control variables
(i.e., cognitive ability, task knowledge and gender), where scores for each variable were
assigned to groups by combining individual attribute values in a somewhat arbitrary
fashion. Given the relatively strong relationships between composition, process and
outcome variables observed in this study (as well as others), future research would benefit
from the development of composition and cross-level theories and the use of more
sophisticated analytical techniques in understanding how these individual difi‘erence
variables operate in group settings.

With respect to improving the reliability of group process measures, there are at
least two strategies. One strategy would involve abandoning the use of observer ratings
and the videotaping group discussion in conjunction with an attempt to develop reliable

paper-and-pencil measures that could be completed by the group. While this strategy

1 36
would involve considerably less work in the long run compared to reliable observer
ratings, it brings to the foreground complicated issues related to method bias, assessing
intragroup agreement and the appropriateness of using aggregated individual perceptions
to represent the group as a whole. A second strategy for improving measurement
reliability would involve the continued use of observer ratings along with an extensive
analysis of the experimental task in the hope of identifying all possible behaviors that
constitute instances of each of the various measures. If all (or even most) potential
behaviors could be identified for each process construct within the confines of a particular
task, the "rating" process could be reduced to a checklist procedure, with an
accompanying shift in focus away from examples and interpretation towards
comprehensiveness and recognition. In implementing this strategy, it almost goes
without saying that it is necessary to assemble (and retain) committed raters who are
given standardized training involving practice, "true score" feedback, refresher sessions
and incentives for accurate coding.

Although a direct comparison of the two alternative models of ill-structured group
decision making would be helpful, future research should also consider including other
group composition variables not examined in this study. As evident fi'om r-square value
generated by the regression analysis associated with Hypothesis 1, a great deal of the
variance in group performance was not explained by the variables in the analysis. There
is a fair amount of evidence that the gender composition of groups will affect the
dynamics of group interaction (Moreland & Levine, 1992). It may also be the case that
various combinations of one or more Big 5 personality constructs such as extroversion,
agreeableness or neuroticism at the individual-level may give groups a distinctive
personality "profile" which affects the manner in which information is processed in the

group.

137

With regard to interventions in group process designed to improve group
performance, the DI manipulations used in this study were not successful. One potential
problem that might be addressed in future research would be to modify the experimental
procedures used in this study so that "plan" and "counterplan" are created in a serial
fashion, with one building on the other, so that both sides have an opportunity to study
the other plan before the dialectical process begins. This would negate the need for
criticisms and questions to be generated "on-line," and might substantially improve the
quality of the dialectical process (as well as increasing the chances of getting
"diametrically opposed" plans.

Also, with regard to future research on prescriptive interventions, it may be that
DI and DA would be more efl‘ective when employed using multiple individuals
advocating "plan" and "counterplan." In order to keep the number of participants to a
manageable level in this study, it was necessary to limit the assignment of "plan" and
"counterplan" to one group member each. It may be that the intellectual stimulation and
moral support provided by a partner would allow for a more in-depth debate when plan
and cormterplan are debated. Future research might address this issue by comparing DI
with one-person roles and DI with multi-person roles.

Further, it should be noted that the interventions used in this study constitute only
a small sample of the difl‘erent ways in which controversy can be increased in decision
making groups through the use of some intervention. To begin with, to the extent that
implementation problems are suggested by the lack of efl‘ects for the DI methods used in
this study, it may be possible to use a functionally-equivalent controversy-inducing
technique that is easier to implement on a procedural basis. For instance, Devil's
Advocacy involves only one role assignment and does not necessitate the creation of

plans or cormterplans or a rigid Sequence of events during group discussion. Future

138
research might employ DA in situations where participant confusion is expected with the
more structured and involved DI process.

Finally, there is a clear need to expand the parameters of the research setting used
in identifying and testing models of group decision making in ill-structured environments.

The current study was based on undergraduate students with little or no history or
collective future who tended to have no managerial experience and who were not given
the opportunity to develop any of these characteristics because of the "one-sho " nature of
the study. Future research attempting to identify a model of ill-structured group decision
making would benefit most from using intact groups of managers who know and interact
with one another on a regular basis in conjunction with longitudinal designs that allow
learning and development. After such a model has been developed, it would be helpful to
demonstrate the model's explanatory power using multiple tasks that would rule out the
possibility of task-specific findings.

99.11%

This study identified two types of process loss that may hamper group decision
making efl’orts: a failure to share information among members and a failure to optimally
use information that is shared. A model was proposed integrating the process findings in
the psychological literature and the prescriptive interventions identified in the
management literature. Despite the presence of several factors which made this study a
conservative test of the hypotheses generated by the model, the pattern of findings was
not inconsistent with the predictions of the model and provided some support for the
incremental validity of process-related factors in explaining group performance. On the
other hand, the expected advantages of the dialectical inquiry conditions were not found.

Future research attempting to identify a model of group decision making in ill-structured

139
contexts would greatly benefit from the longitudinal study of intact, ongoing decision
making groups and the replication of findings across various types of ill-structured task

environments.

APPENDICES

APPENDIX A

APPENDIX A

SOUTHEAST AIRLINES, INC.:
A BUSINESS SINIULATION IN THE AIRLINE INDUSTRY

1. OVERVIEW

Welcome to SouthEast Airlines, Inc. In this simulation, you and three other individuals
play the role of a top management team charged with creating a strategic plan for SouthEast
Airlines for the upcoming fiscal year. Each member of your group will be assigned to one of the
following positions: (1) Vice Presideng, Flight Operations. (2) Vice President. F inappg, (3)3133;
President, Marketing, or (4) Vice Presideng, Indam Analysis.

In order to develop a strategic plan, your group will be provided with information on last
year’s operations and what can be expected in the future. You will each receive information that
corresponds to your position. You will be asked to familiarize yourself with this information and
apply it dming group discussion. Like real-world business operations, this simulation is
relatively complex and will be confusing at times. However, by the time you are finished, it will
make sense. Do the best you can as a group, and remember - it's just money!

YOUR OBJECT AS A TEAM IS TO IDENTIFY A PLAN THAT WILL EARN THE
MOST PROFIT FOR " SOUTHEAST AIRLINES." This plan is the final product of your efforts.

H. SEQUENCE
There are two phases in the creation of a strategic plan:

1. Individual Preparation Phase (60 minutes)
2. Group Planning Phase (75 minutes)

In the Individual Preparation Phase, you will be given a packet of information relevant
to your position in the company. Use the 60 minutes to become thoroughly familiar with it.
During group discussion, you will not have time to go back and "learn" this material. It will be
EXTREMELY HELPFUL to your group if you use the time allowed to prepare yourself wellll

In the Group Planning Phase, your group will be reassembled for the purpose of
reaching agreement on a final plan. In a few minutes, you will be given special instructions for
how to proceed as a group in this phase.

AQAW, YOUR OBJECT AS A GROUP IS TO COME UP WITH A PLAN THAT RESULTS
IN MAXIMUM PROFIT FOR SOUTHEAST AIRLINES!

140

141
APPENDIX A

III. HOW THE SIMULATION WORKS

Your airline is based in Atlanta, GA. There are a number of other cities represented in
this simulation. Your firm generates revenue by providing airline transportation between Atlanta
and these other cities, but this also incurs various costs. Your team will be trying to come up
with a plan that maximizes total revenue and minimizes costs. Profit is determined as follows:

Profit = Total Revenue + Ca§_h - Costs - Debts

Total Revenue = sum of Route Revenues

Route Revenue = (Market Share x Passenger Demand x Fare) per route

Cash = Invested Cash x Interest Rate

Costs = Aviation Fuel + Facilities/Equipment + Flight Staff + Ground Staff +
Maintenance + Marketing + Purchases + Loan Repayment + Finance Charges

Debts = Balance of outstanding loans

Setting aside Cash and Debts for the moment, it can be seen from the profit formula that
the more Total Revenue you generate, the greater your profit. At the same time, the more Costs
you incur, the lower your profit. Thus, you should strive for a plan that generates as much Total
Revenue as possible while minimizing Costs.

LL'axraLiLingTotal Revenue and Maxim ' ’ ' Costs

 

Figure 1 provides a graphic display of the factors that affect profit on any given route.
Arrows in the diagram depict causal relationships between variables. When there is an arrow
between two variables, the variable to the left at least partially determines the level/amount of
the variable to the right of the arrow. The variables and their relationships are explained in detail
in the information provided to you group.

In general, the five “little” variables at the left are the factors over which you and your
group have the most control. In most cases, you can simply choose the values of these variables
(for example, deciding to offer five Daily Flights on a route). Once you have made choices with
regard to these five variables, they affect other variables further to the right. Note that Total
Revenue is simply the sum of all the Route Revenues, and Route Revenues are greatest when
Fare Price is high, Passenger Demand is high, and Market Share is high. Although Fare Price
and Passenger Demand are relatively straightforward, Market Share is a complex variable
afi'ected by all five of the “little” variables. Further, the five “little” variables affect Costs by
determining the number of planes used on a route, the amount of aviation fuel consumed and the

142
APPENDIX A

number of staff required to operate/ service the route. Generally, as you do things to increase
Market Share, you also increase Costs to some extent. The goal of the simulation is finding
routes where you can: (1) Charge a high Fare Price, (2) Expect high Passenger Demand, (3)
Establish a good Market Share and (4) Avoid excessive Costs. In a nutshell, this is how you
succeed in “SouthEast.”

IV. COMPLETING THE STRATEGIC PLANNING DOCUDIENT

A completed Strategic Planning Document is the final product of your group effort. All
actions you desire to implement as a group MUST be specified on the Strategic Planning
Document. To complete a Strategic Planning Document, you MUST do at least five things:

1). 1""“SELECT ROUTES FOR AIRLINE SERVICE“r
and for each selected route:

2). Decide the # and TYPE of AIRCRAFT to use

3). Decide the # of DAILY FLIGHTS to offer

4). Decide on the PRICE of the FARE/TICKET

5). Decide on the # of FLIGHT STAFF to have on each flight

In addition to these things which you must do, there are a number of other activities
which you MAY choose to do if you wish. The information you will be given covers these
actions in more detail. In general, you can indicate your desire to conduct these actions by noting
your intentions to do so in the appropriate place on the third page of the Strategic Planning
Document. For now, these actions are:

1. Invest Cash: You have $24.3 million in cash and short-term investments. Should you
decide to invest some or all of this money, it draws interest at the Investment Interest
rate.

2. Market/Advertise: You can spend money on marketing campaigns by advertising
SouthEast’s service in one or more cities using various media (e.g., TV, radio,
newspaper, billboard) if you desire. This increases Market Share on routes that connect
with the city where marketing/advertising is being conducted.

3. Purchase Aircraft: You can buy additional aircraft and use these aircraft as you would
the ones you already own.

143
APPENDIX A

Sell Aircraft: You can sell aircraft that you do not intend to use. This provides you
with more cash.

Pay off loans at an accelerated rate: SouthEast has two outstanding loans. Although
there is a certain amount you are required to re-pay next year, you can accelerate
repayment (i.e., pay more than the minimum) and reduce your finance charges. This
reduces Costs.

144

APPENDIX A

STRATEGIC PLANNING DOCUIVIENT

RULES FOR FILLING OUT THE STRATEGIC PLANNING DOCUMENT:

a).

b).

c).

d).

e).

YOU CAN ONLY HAVE ONE TYPE OF PLANE ON A ROUTE.

UNLESS OTHERWISE STATED, ONE AIRCRAFT CAN MAKE TWO
ROUNDTRIP FLIGHTS PER DAY.

FARE PRICES MUST BE WITHIN $100 OF THE AVERAGE FARE FOR
THE ROUTE AND MUST BE IN MULTIPLES OF $25 (e.g., $125, $250,
$475).

ALL ROUTES MUST BE NONSTOP BETWEEN ATLANTA AND SOME
OTHER CITY.

THE PLANNBNIG DOCUMENT MUST BE SIGNED BY ALL TEAM
MEMBERS

NOTE: FAILURE TO ABIDE BY THESE CONSTRAINTS WILL NULLIFY YOUR

PLAN!

 

All members MUST provide a signature in one of the spaces provided below for the plan to

 

 

 

 

be valid:
(VP, Flight Operations) (VP, Marketing)
(VP, Finance) (VP, Industry Analysis)

 

 

Group #

Date

 

145

APPENDIX A

Strategic Planning Document

 

 

Group #: Date:
Route: Aircraft Aircraft Daily Flight Fare“ ,
ATLANTA- Type“ #14 Flights“ Staff

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

** See restrictions

 

146

APPENDIX A

Strategic Planning Document

Marketing Efforts: Place an “X” where you wish to designate marketing efforts

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

City Television Radio Newspaper Billboard
Amount Cash Invested: $
Extra Loan Repayment: LOAN A: $
LOAN B: $
Aircraft Purchased: Aircraft Sold:
Type # Type #
A-300 A-300
DC-9 DC-9
B-757 B-757
L-101 l L-lOl l
B-727 B-727
DC-8 DC-8
B-747 B-747
DC-10 DC-10

 

 

 

 

 

 

 

 

 

147
APPENDIX A

SOUTHEAST AIRLINES YEAR-END REPORT

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

TO: ALL VICE PRESIDENTS
FROM: BOARD OF DIRECTORS

00000000000000000000000000000000000000000000000000000000

Note from the BOARD:

In the 1994-1995 fiscal year, SouthEast Airlines generated a profit of $11.78 million. This is
considerably lower than the average profit over the last five years. We are somewhat concerned
with this "slide," and hope that your efforts might reverse this trend.

You should find the information contained in this report useful for your meeting. We want you
to consider all possibilities. As you are aware, there really wasn't a formal strategic planning
process last year, and that hurt us.

THE PURPOSE OF YOUR MEETING SHOULD BE TO GENERATE A PLAN THAT
WILL BRING IN MORE REVENUE THAN LAST YEAR'S OPERATIONSII

148
APPENDIX A

SOUTHEAST YEAR-END REPORT

Overview

SouthEast Airlines, Inc. is headquartered in Atlanta with primary operations centered at
the Atlanta International Airport. Currently, SouthEast operates 39 aircraft and provides services
to 11 other cities in the United States. At the end of the last fiscal year, we employed 569
personnel classified as flight staff and 720 personnel classified as ground staff.

Histogy

SouthEast Airlines was founded in 1952 as a small regional airline intended to provide
air transportation to passengers traveling in the Deep South. Initially, SouthEast served four
cities
- Dallas, New Orleans, Miami, and Nashville. SouthEast grew during the 1960s and expanded
service to four more cities in the southeastern United States (Raleigh-Dmham, Tampa,
Louisville, Memphis). In the 1970s, service was extended to a number of cities outside the Deep

South (Chicago, New York, Los Angeles).

The company entered a period of financial difficulty in the early 19805 due to the
national economic recession and federal deregulation of the airline industry. In 1982, SouthEast
Airlines lost money for the first time in its history, and did so again in 1983 and 1984. The
situation has tinned around since 1984, with annual profits ranging from a low of $28.3 million
in 1985 to a high of $139.1 million in 1992. However, during the last fiscal year, profits fell to
their lowest level since 1984—1985 (approximately $12 million). There are a variety of reasons
for this decrease, but further decline is not acceptable.

149

APPENDIX A

SOUTHEAST YEAR-END REPORT

Table 1. Flight Operations Dag

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Route Aircraft # Aircraft Flight Daily
Type Assigned Staff Flights

Memphis B-757 3 6 5
Louisville DC-9 3 6 6
Nashville DC-9 3 6 6
Raleigh-Durham A-300 4 4 7
Dallas-Ft. Worth DC-8 4 10 7
New Orleans B-757 3 7 5
Tampa L-lOl l 3 8 6
Miami B-727 5 8 9
Chicago B-747 4 10 7
New York B-727 4 9 8
Los Angeles D101] 3 12 6
Totals/Ave. - 39 - 72

 

 

Table 1 Notes:

Aircraft Type refers to the type of aircraft used on the route. Our policy is to use only
one type of aircraft on a particular route so customers know what to expect.

# Aircraft Assigned is the number of aircraft employed on the route. FAA regulations
prevent planes from making more than two roundtrip flights per day, necessitating a 1:2
ratio of aircraft to flights on all routes.

Flight Stafi' represents the number of flight attendants assigned to work EACH flight.

Daily Flights represents the number of flights offered on a given route per day..

Table 2. Revenue Information

150

APPENDIX A

SOUTHEAST YEAR-END REPORT

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Route Previous Ave. Our Pass. Market Revenue
Competition Fare Fare Demand Share (millions)
Memphis Low $250 $200 200,000 43% 17.20
Louisville Low $175 $250 125,000 35% 10.94
Nashville Moderate $200 $225 175,000 30% 11.81
Raleigh- Moderate $175 $125 200,000 35% 8.75
Durham
Dallas- Heavy $275 $325 500,000 24% 39.00
Ft. Worth
New Orleans Heavy $250 $300 200,000 8% 4.80
Tampa Heavy $225 $250 225,000 16% 9.00
Miami V. Heavy $250 $300 525,000 8% 12.60
Chicago V. Heavy $300 $350 600,000 16% 33.60
New York V. Heavy $300 $300 550,000 15% 24.75
Los Angeles V. Heavy $400 $425 $75,000 8% 19.55
Totals/Ave. - $254.55 $277.27 352,273 22% 192.00
1%

Previous Competition is an indication of how many competing airlines also offer flight
services on the route in question.

Ave. Fare is the price of the “average” fare offered by our competitors for the route.

Our Fare represents the price we charged for a roundtrip ticket last year.

Passenger Demand represents the number of people traveling roundtrip between Atlanta

and the various other cities over the course of the last fiscal year.

Market Share represents the percentage of the Passenger Demand that used Southeast.

Revenue is how much money we generated on the route last year (in millions).

 

SOUTHEAST YEAR-END REPORT

Balance Sheet:

Liquid Assets: $24.3 million in CASH and short-term investments

151

APPENDIX A

W: $23.5 million in outstanding loans

Last Year's Passenger Revenue: $ 192.00 million
Last Year's Cost: $ 180.22 million

Last Year’s Operating Profit: $ 11.78 million

Table 3. Cost Brea_kdown

 

 

 

 

 

 

 

 

 

 

 

 

 

Type Cost °/o of Total
Aviation Fuel 65.91 M 36.6
Facilities/Equipment 48.00 M 26.6
Ground Staff 28.80 M 16.0
Flight Staff 25.61 M 14.2
Maintenance 5.85 M 3.2
Loan Repayment 2.75 M 1.5
Finance Charges 2.30 M 1.3
Advertising/Marketing 1.00 M 0.6
Total 180.22 M 100.0

 

Note: Cost is in millions of $

 

152
APPENDIX A
SOUTHEAST YEAR-END REPORT

Table 4. Profit Information by Route (in millions)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Route Route Route Profit Profit

Revenue Cost (millions) Ratio
Memphis $17.20 $10.36 $6.84 .66
Louisville $10.94 $11.76 - $ .82 -.07
Nashville $11.81 $10.97 $ .84 .08
Raleigh-Durham $ 8.75 $10.86 - $2.11 -. 19
Dallas-Ft. Worth $39.00 $17.35 $21.65 1.25
New Orleans $ 4.80 $10.99 - $6.19 -.56
Tampa $ 9.00 $14.75 - $5.75 -.39
Miami $12.60 $20.10 - $7.50 -.37
Chicago $33.60 $19.77 $13.83 .70
New York $24.75 $20.55 $4.20 .20
Los Angeles $19.55 $32.69 - $13.14 -.40
Totals/Ave. $192.00 $180.22 $11.78 .08

Table 4 Notes:

Route Revenue is the total $ income generated by the route last year (in millions).
Route cost is the total $ cost to operate the route last year (in millions).

Profit is (Route Revenue - Route Costs)

Profit Ratio is simply Profit/Route Costs and represents a standardized measure of
Rettun On Investment (ROI). Note that a negative profit ratio indicates the loss of

money on a route, 0.00 is the “break even” point, and a ratio of 1.00 would be equivalent
to earning twice as much revenue on a route as it cost to operate (i.e., 100% R01).

153
APPENDIX A

MEMO

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

TO: Vice President, Flight Operations
FROM: Flight Operation Staff

This memo is in response to your inquiry about route efficiency. As you know, a number of our
routes lost money last year, and a primary reason was because of poor decisions concerning the
allocation of aircraft to routes, the number of daily flights to offer, and the number of flight staff
to put on each flight. We have prepared an analysis of costs associated with flight operations on
orn- various routes, and provided some conclusions and recommendations for you to consider.
Since fuel costs are such a big percentage of our total costs, we thought you might want to know
how we calculated fuel costs for the routes. We did it using this formula:

Route Fuel Costs = (Cost/Flight * Daily Flights * 365 days)
where Cost/Flight = (Cost/Mile * Roundtrip Distance)
For example, the aircraft flying the Memphis route have a Cost/Mile of $1.50 and the
roundtrip distance is 600 miles. We multiplied these two values together to get Cost/F light

($900), and then multiplied $900 Cost/F light by 5 Flights/Day and 365 Days to get the Annual
Fuel Cost for the Route. This information is presented for each route in the following table.

154

APPENDIX A

Summary: Annual Fuel Costs by Route

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Route Cost/ Rd. Trp. Cost/ Daily Annual
Mile Distance Flight Flights Fuel Cost
Memphis $1.50 600 $900 5 $1.643 M
Louisville $1.80 600 $1080 6 $2.365 M
Nashville $1.80 400 $720 6 $1.577 M
Raleigh-Durham $1.00 500 $500 7 $1.278 M
Dallas-Fort Worth $2.30 1000 $2300 7 $5.877 M
New Orleans $1.50 800 $1200 5 $2.190 M
Tampa $2.75 800 $2200 6 $4.818 M
Miami $2.10 1100 $2310 9 $7.588 M
Chicago $2.50 1300 $3250 7 $8.304 M
New York $2.10 1400 $2940 8 $8.585 M
Los Angeles $2.75 3600 $9900 6 $21.681 M
Totals/Ave. $1.97 1100 $2445 72 $65.91 M
Nagaag

Cost/Mile measures the cost in dollars for an aircraft to fly one mile. Each aircraft has a
Cost/Mile rating from $1.00 - $2.75. This value is multiplied by the distance in miles to
determine the cost of one aircraft making a single, one-way flight on a given route.

Rd. T rp. Distance is the roundtrip distance in miles from Atlanta to the various cities.
Cost/Flight is simply (Cost/Mile * Rd. Trp. Distance).

Daily Flights is the number of daily roundtrip flights on the route.

Annual Fuel Costs is the total amount in millions that it cost to ptu'chase aviation fuel for
planes operating on the route over the course of the last fiscal year.

155
APPENDIX A

Summary: Aircraft Characteristics & Cost Information

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Type Accom. Flight Staff Cost/Mile Purchase
Class Lim. Cost (mil).
A-300 E 4 $1.00 $1.0 M
DC-9 D 8 $1.80 $2.0 M
B-757 D 7 $1.50 $2.5 M
L-lOll C 12 $2.75 $3.0 M
B-727 C 9 $2.10 $4.5 M
DC-8 B 11 $2.30 $6.0 M
B-747 B 15 $2.50 $6.0 M
DC-10 A 12 $2.40 $8.0 M
15913

Aircraft Type is the formal FAA designation for the aircraft used on the route.

Accommodations Class refers to the ergonomic, user-friendly aspects of an aircraft's
design. There are 5 classes of Aircraft Accommodations, ranging from "E" (poor) to "A"
(excellent).

Flight Stafl' Limit refers to the maximum number of Flight Staff that can effectively serve
aboard the aircraft. Each flight requires a separate crew.

Cost/Mile represents the cost in dollars for flying the plane one air-mile.

Pro'chase Cost refers to the price associated with the pch of one new aircraft of the
type indicated (IN MILLIONS OF DOLLARS).

156
APPENDIX A
Market Share Modifiers

We have prepared a few tables to show how changing these values can influence our Market
Share on a given route. To use these tables, figure out which column to use by estimating the
Competition and the level of the modifying variable (e.g.., Daily Flights). Then, look down the
various rows and compare the various options in terms of their effect on Market Share. In
general, lots of daily flights, excellent accommodations and lots of flight staff translate into big
market shares.

Modifier #1: Convenience (Daily Fligata)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

COMPETITION
Daily Low- Heavy-
Flights Moderate Very Heavy
1 -12% -25%
2 -10% -l6%
3 -8% -l 1%
4 -5% -7%
5 -2% -4%
6 0% -2%
7 +2% 0%
8 +4% 0%
9 +6% +1%
10 +8% +3%
1 1 +9% +5%
12 +10% +7%
13 +1 1% +8
14 +1 1% +9%

 

 

 

 

 

Note: Values in table are ADJUSTMENTS to Market Share

157

APPENDIX A

Modifier #2: Comfort (Accommoaaaions)

 

 

 

 

 

 

 

 

 

 

 

 

Roundtrip Distance
Factors <1500 miles 1501+ miles
Competition Low or Heavy or Low or Heavy or
Accommodations Mod. V. Heavy Mod. V. Heavy
"E" Class -2% -10% -20% -3 5%
"D" Class 0% -5% -10% -20%
"C" Class 0% 0% -1% -7%
"B" Class +2% +5% +7% +5%
"A" Class +3% +7% +12% +9%

 

Note: Values in table are ADJUSTMENTS to Market Share

 

158

APPENDIX A

Modifier #3: In-Flight Service (Flight Staff)

To estimate the modification to Market Share for the number of Flight Staff, we use a
different rule of thumb: Take the number of Flight Staff on the route and subtract the value for
the Average Flight Staff for competitors. Then, multiply this number by one or two, depending
on the competition (1= Low or Moderate Competition, 2 = Heavy or Very Heavy Competition).
The resulting number is the adjustment to existing market share. Here is the formula:

 

(Flight Staff- Average Flight Staff) * 1 or 2“

”Low OR Moderate Competition = 1
”Heavy OR Very Heavy Competition = 2

"Flight Staff" refers to the number of Flight Staff you decide to put offer on a given

route.
"Average Flight Staff" refers to what your competition on the route is offering.

Example:

You want to know what the effect on Market Share would be if you use put 13 Flight
Attendants on each flight on a route where the Average Flight Staff value = 12 and
where there is Very Heavy Competition (Multiplier = 2). The modification would be:
(13-12) * 2 = 2. Thus, if you were to use 13 Flight Attendants for flights on this route,
there would be a “+2” modification to our Market Share.

159

APPENDIX A

Analysis and Conclusions.

1.

There is a trade-off between fuel-efficiency and passenger comfort. We need fuel-
efficient planes on the longer routes, but these aircraft need to have good
accommodations. Some of our longer routes definitely need new planes!

2. We can drastically affect Market Share by altering the number of Daily Flights, type of
aircraft (i.e., accommodations) and the number of Flight Staff assigned to a route -
possibly increasing Market Share by up to 40%!

3. Many of the routes where we are losing money could be made more profitable by simply
adding a few more flights, getting better aircraft (including more fuel efficient ones on
the long routes) and/or adding a few Flight Staff to each flight.

Recommendations.

1. Juggle existing aircraft assignments so as to minimize fuel costs and maximize market
share bonuses for good Accommodations.

2. Sell some of the old fuel-inefficient aircraft and pruchase new ones that have (1) decent
fuel-efl‘iciency and (2) good Accommodations.

3. Adjust Daily Flights and Flight Staffs to maximize market share bonuses for

Convenience and In-F light Service.

160
APPENDIX A
MEMO

TO: Vice President, Finance

FROM: Finance Department Staff

In keeping with your request to put together some information for your upcoming planning
meeting with the other Vice Presidents, we have compiled the following material which we hope
you will find useful. We begin with a description of the various kinds of costs we incur in our
operations, then provide a table displaying our unit costs in each area last year and projected
costs for next year. We conclude with some options for reducing costs in next year’s plan.

Description of Costs
Aviation Fuel. Costs associated with the purchase of aviation fuel for our aircraft.

Facilities Costs associated with renting hangars, offices, storage space,
equipment, etc., on every route we service. We currently have
12 domestic facilities (counting Atlanta).

Flight Staff. Costs associated with the employment of flying crews and flight
attendants. fle total number of Flight Staff on agiven route is

equal to Flight Staff "' Daily Flights (each flight requires a separate
crew). Last year, we employed 579 F light Staff.

Ground Staff. Costs associated with the employment of all non-flight personnel. It works out
that we need 10 gzound staff per fligm, so the total number of Ground Staff can
be determined by multiplying Total Daily Flights by 10. Last year, we employed

720 Ground Staff.
Maintenance. Costs associated with routine maintenance, inspection, and repair of our aircraft
fleet.
Marketing. Costs associated with advertising campaigns in one or more cities using

one or more different media.
Purchase. Costs associated with the purchase of new aircraft.

Loan Repayment. Payments made on the balance of the principal for SouthEast’s two long-
term loans.

Finance Charges. Interest paid on the outstanding balance of SouthEast’s two
long-term loans.

 

161
APPENDIX A

Previous and Projected Costs

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Item Last Year Projected Next Yr.
Aviation Fuel —- -—
Ground Staff personnel @ $40,000 $42,000
Flight Staff personnel @ $45,000 $49,000
Maintain Domestic Facility @ $4,000,000 $4,250,000
Start-up Domestic Facility @ -- $5,000,000
Start-up Foreign Facility @ — $7,000,000
Aircraft Maintenance @ $150,000 $160,000
Marketing —— .....
Loan Repayment (A) $1,500,000 $1,500,000
Loan Repayment (B) $1,250,000 $1,250,000
Finance Charges (A) $633,750 $536,250
Finance Charges (B) $1,663,750 $1,526,250
Loan LD. Outstanding Interest Rate

Principal
001-91 (A) 9,000,000 6.5%
002-94 (B) 14,500,000 11.0%

 

 

 

 

 

We are required by our loan agreements to pay the minimum listed in the table on the previous
page (minimum = “projected”). However, we can save money by paying off these loans faster.
To do this, all we need to do is indicate how much extra we want to pay on the second page of
the Strategic Planning Document. Finance Charges are based on the average monthly balance,
much like a credit card. The more you pay on the outstanding principal, the less the Finance
Charges will be.

 

162
APPENDIX A
Conclusions

1. The bulk of our costs come from aviation fuel, leasing facilities/equipment and paying for
staff (both ground and flight). These are primary areas for cutting costs.

2. It takes 10 Ground Staff for each and every flight we offer. Cutting back on Daily Flights
where possible could result in big savings.

3. Each flight utilizes its own separate flight staff. Therefore, adding one flight staff to each
flight on a route can result in adding up to 14 Flight Staff personnel - depending on the number
of Daily Flights. Thus, for routes with many flights, adding Flight Staff can get very expensive!

Recommendations.

1. Eliminate inefficient routes — cutting routes which lost money will automatically result
in additional profit.

2. Buy planes that are more fuel efficient.

3. Pay off existing loans at an accelerated rate and reduce finance charges.

 

163
APPENDIX A

MEMO

TO: Vice President, Marketing
FROM: Marketing Staff

re: Fare prices and advertising information

We just got the analysis back from the big customer satisfaction survey we did last year.
As you know, the basic formula we use to calculate revenue on each route is as follows:

Passenger Revenue = Passenger Demand * Market Share * Fare

Simply put, the money we make on each of our routes is equal to the number of people
who fly SouthEast multiplied by the price of their fare. For a given route, the number of people
who fly SouthEast is equal to the total number of people traveling (Passenger Demand)
multiplied by our Market Share for the route.

We make the most money when we offer services between big cities, when we have high
Market Share and when our fares our high. Unfortunately, the inverse relationship between Price
and Demand results in low Market Share when prices are high, and vice versa. Therefore,
picking the optimal price is a bit tricky.

The data from our survey indicate that the two primary factors affecting how Fare Price
affects Market Share are: (1) Deviation from Average Fare and (2) the Level of Competition on
the route.

On the next page, we provide a very important table for determining Market Share.
With the table, we can figure out what the "optimum" price of our fares is for each route.

164
APPENDIX A
_R_elations_hia) between Fae Price and Market Shara
To use table:
1. Choose a possible price
2. Determine this price's deviation from the route's Average Fare

3. Determine level of Competition on route
3. Cross-index row and column to see what Market Share would be

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Competition

Deviation from Low Moderate Heavy Very Heavy
Average Fare

$ 100 under ave. 53% 44% 35% 26%
$75 under ave. 48% 39% 31% 25%
$50 under ave. 44% 35% 28% 24%
$25 under ave. 41% 32% 25% 21%
Same as Ave. 40% 30% 22% 17%
$25 over ave. 39% 28% 19% 13%
$50 over ave. 37% 24% 16% 10%
$75 over ave. 34% 20% 12% 9%
$100 over ave. 30% 15% 9% 8%

Comments:

The "Same as Ave." row in the table reflects our Market Share if we set the price of our
Fare equal to the average price of our competitors.

The optimum price for any particular route depends on the level of competition and the
value of the average fare. This can be determined by calculating “expected value.”

To calculate an "expected value," multiply the price of a potential fare (e.g., $150 or
$400) by the market share it would have (e.g., .22 or .45). The higher the expected
value, the more money we will make if we use that price. To compare several price
options, do the same for each potential price and compare their expected values. The
fare with the highest expected value is the optimum choice for pricing.

165
APPENDIX A

Marketing Modifiers

Market share on the various routes can be increased by advertising in the media.
Basically, by advertising SouthEast's services in the various cities we serve, we can increase our
market share over and above what it would be based simply on the price of the fare. Last year,

we only maaketed in Atlanta, and we confined our advertising to outdoor billboards. The
following table shows the modification to Market Share when advertising using various media in

cities on our routes:

 

 

 

 

 

 

 

 

 

 

 

 

Media Cost/City Effect/Route - Effect/Route -
(Low/Moderate (Heavy/V. Heavy
Competition) Competition)

Television $5,000,000 +3-4% +2-5%

Radio $2,000,000 +2-3% +2-3%

Newspaper $1,000,000 +2% +1-3%

Billboard $500,000 + 1% + 1-2%

NONE O 0% -3%

HEELS;

The cost paid to advertise in a city affects all routes into/out of the city. For instance, if
we advertise in Atlanta, we pay ONE cost but get the advertising bonus for ALL routes.
Ifwe advertise in cities other than Atlanta, the bonus will apply only to the route
connecting Atlanta and the city in question. However, we can market in Atlanta as well
as other cities and the effect is cumulative.

The result of using more than one medium is NOT the sum of the individual bonuses.
The combined adjustment will be less than the sum of the individual bonuses (approx.
75%) because of redundancy in media coverage of the population.

166

APPENDIX A

_A_nalysiL and Conclusions:

1.

We should raise our fares when facing Low or Moderate Competition and lower our
fares when Competition is Heavy or Very Heavy.

2. The value of a market share percentage point depends on the Passenger Demand for the
route - an extra point on busy routes means a lot more than an extra point on a
secondary route.

3. Some of our prices are really out of whack! Simply by finding the optimum price, we
could increase Market Share on some routes by 15%!

4. Relative to their cost, marketing efforts are probably worth the cost - especially in
Atlanta. By using all four media, we can increase our Market Share on each route by
5-6%!

5. If we don’t market, we will get hit hard on routes with “Heavy” or “Very Heavy”
competition.

Recommendations:

1. Revise existing price structtne to maximize expected value.

2. Heavily advertise in Atlanta using multiple media.

3. Consider advertising in a few key metropolitan areas like Chicago or LA.

167
APPENDIX A

MEMO

TO: Vice President, Industry Analysis
FROM: Industry Analysis Staff

After scouring the information in the federal publications, we've finally got the
information you asked for: Estimates for Passenger Demand and Competition for SouthEast's
existing routes and a number of prospective routes. We lay out this information on the next two
pages, then provide an analysis and some recommendations for you to consider.

As you know, the basic formula we use to calculate revenue for each route is as follows:

Passenger Revenue = Market Share * Passenger Demand * Fare

We don't have the complete analysis on Price-Market Share relationships, but we do
know what Market Share we can expect if we adopt the Average Route Fare as our own:

Low Competition: 40%
Moderate Competition: 30%
Heavy Competition: 22%
Very Heavy Competition: 17%

Using the formula above and plugging in these average figures, it should be possible to
determine some new routes that bring in more money than some of the existing routes.

A NOTE ON COMPETITION: It is extremely difficult to estimate what sort of competition we
will face on routes next year. The number of competitors could change, fare wars could break
out, etc. As a result, we have estimated the probability of facing each level of competition. For
some routes, we can be pretty certain what the competition will be like. For others, it’s anyone’s

guess.

168

 

 

 

 

 

 

 

 

 

 

 

 

APPENDIX A
Existing Routes: ATLANTA—
Route Demand Rd. Trp. Competition Ave. Ave.
Distance Likelihood Fare Flight Staff
Memphis 195-205,000 600 70-15-15-0 $250 6
Louisville 100-150,000 600 75-15-10-0 $175 6
Nashville 170-210,000 400 60-30-10-0 $200 5
Raleigh- ZOO-225,000 500 50-35-15-0 $175 5
Durham
Dallas- 425-475,000 1,000 5-20-60-15 $275 9
Ft. Worth
New Orleans 190-220,000 800 15-35-35-15 $250 7
Tampa 215-245,000 800 5-30-50-15 $225 9
Miami 500-530,000 1,100 0-15-65-30 $250 10
Chicago 605-625,000 1,300 0-5-15-80 $300 10
New York 560-5 80,000 1,400 0-10-15-75 $300 1 1
Los Angeles 585-610,000 3,600 0-10-20-70 $400 12

 

 

 

 

 

 

 

Note: "Demand" = Expected Passenger Demand for upcoming year IN THOUSANDS

Note: "Competition Likelihood" = % chance that competition on the route in the upcoming
year will be low-moderate-heafl-vgy beau, respectively.

 

APPENDIX A

 

 

 

 

 

 

 

 

Potential Routes: ATLANTA-
Route Demand Rd. Trp. Competition Ave. Ave.
Distance Likelihood Fare Flight Staff
Madrid 85-90,000 8,000 70-15-15-0 $875 12
Paris 75-125,000 9,000 55-20-20-5 $900 1 1
London ISO-200,000 9,000 50-30-10-10 $850 12
Sao Paulo-Rio 80-100,000 12,000 85- 1 5-0-0 $1200 10
Mexico City 110-175,000 2,800 40-30-20-10 $500 1 1
Cancun 90-120,000 3,000 55-30-10-5 $500 13
Virgin Islands 85-145,000 4,000 45-35-15-5 $550 1 l

 

 

 

 

 

 

 

Note: "Demand" = Expected Passenger Demand for upcoming year IN THOUSANDS

Note: "Competition Likelihood" = % chance that competition on the route in the upcoming
year will be low-moderate-heag-very heayy, respectively.

 

170

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

APPENDIX A
Potential Routes: ATLANTA-
Route Demand Rd. Trp. Competition Ave. A. Flight
Distance Likelihood Fare Staff

Seattle 305-350,000 4,500 50-30-15-5 450 12
Minneapolis- 25 5-295,000 1,800 65-30-5-0 325 1 1
St. Paul

Cincinnati 265-295 ,000 700 60-20-15-5 250 7
Indianapolis 245-260,000 800 75-20-5-0 225 9
San Francisco 395-470,000 4,200 40-20-20-20 425 11
Denver 420-500,000 2,300 20-40-30-10 375 12
St. Louis 260-280,000 900 25-60-10-5 250 6
Buffalo 270-295,000 1,400 35-30-30-5 300 8
Kansas City 245-255,000 1,300 30-55-15-0 300 9
Pittsburgh 385-450,000 1,000 10-25-55-10 275 8
Philadelphia 335-365,000 1,200 15-25-50-10 300 7
Phoenix 395-490,000 3,100 10-65-20-5 375 13
Boston 47 5-520,000 1,800 10-15-30-45 325 9
Detroit SOD-615,000 1,200 0-10-25-65 325 8
IV)Vr(r:shington, 725-800,000 1,000 0-10-20-70 250 9

 

 

 

 

 

 

 

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

Note: "Demand" = Expected Passenger Demand for upcoming year IN THOUSANDS

Note: "Competition Likelihood" = % chance that competition on the route in the upcoming
year will be low-modgate-heavy-very hem, respectively.

 

171

APPENDIX A

Analysis & Conclusions

1. We could earn a great deal more revenue by picking up some longer routes to large
markets.

2. Big markets with low competition are prime prospects - even if we only offer one flight
per day!

3. Looking at the probability of competition for next year, some cities are definitely "safer
bets" than others.

Recommendations

1. DROP SOME OF THE EXISTING ROUTES AND PICK UP SOME PRIME ROUTES
TO BIG CITIES WHERE LOW COMPETITION IS EXPECTED.

APPENDIX B

APPENDIX B

Instructions to Groups in the Consensus-Seeking Condition

Your group will use the Consensus approach in creating a strategic plan. This approach

is based on a thorough group discussion involving all group members during the Group
Discussion phase. Through questioning, discussion and an open exchange of views, the
Consensus approach should result in a better plan than any single group member could produce.
During the Group Discussion phase, everyone should feel free to offer any and all thoughts,
ideas, and recommendations they have. When you have generated a set of ideas that you can all
collectively agree on, your group has reached "consensus." It is not necessary that you each be
completely satisfied with the final plan — you only have to consider it workable.

Here are some gta'delines for achieving consensus:

1.

Present your ideas clearly and logically, specifically noting any recommendations you
have concerning changes that need to be made.

Avoid thinking that someone must win and someone must lose when there is
disagreement. When a deadlock occurs, look for a compromise solution.

Don't change your position simply to avoid conflict and/or speed things up. Similarly,
avoid things like majority voting, tossing a coin, etc., as a means of solving
disagreements. Differences should be reconciled through discussion.

Be cautious if everyone agrees on something without discussion or examination. All
ideas should be thoroughly scrutinized and alternatives should be considered.

When you all agree on a final plan, record the features of the plan on the Strategic
Planning Report form provided, sign it as a group, and notify the experimenter.

Your plan isn't valid and will not be implemented if it isn't signed by all members. In the
event of an invalid plan, last years plan will be implemented again by default without

any changes.

REMEMBER: YOU ONLY HAVE 75 MINUTES TO FINISH! When the timer
goes off, you must hand in what you have completed or your plan will not be invalid.

172

APPENDIX C

APPENDIX C

Instructions to Groups in Traditional Dialectical Inquiry Condition

Your group will use the Competing Plans approach to help you create your strategic
plan. This simple approach is based on having two group members create their own individual
plans and present them to the group. The Vice Presidents of Flight Operations and Industry
Analysis have been randomly selected to create these plans. Note that your respective staffs have
already generated a number of good ideas. At the beginning of the Group Discussion Phase, the
two plans are presented and then each plan is critiqued by the other presenter. The presentations
should get a number of ideas out on the table for further discussion, then the critiques will help to
"weed out" flawed ideas. Ideas and recommendations that survive the critique are more likely to
be good ideas than those that don't. Here is the sequence you should follow to implement the
Competing Plans approach:

At the start of the Group Discussion Ph_as_e_:

1. The VP of Flight Operations presents Plan A (5 minutes or so)
2. The VP of Industry Analysis presents Plan B (5 minutes or so)
3. The VP of Flight Operations critiques Plan B (5 minutes or so)
4. The VP of Industry Analysis critiques Plan A (5 minutes or so)
5. Open Discussion (Remaining 55 minutes or so)

Here are some gg'delines to follow in implemen_t_r_ng' the Competing Plans approach:

 

1). Again, the VP's of Flight Operations and Industry Analysis get to present and critique
plans. The VP's of Finance and Marketing should hold any questions and comments
until after the critiques. After the critiques have been conducted, everyone is free to say

anything.

2). The VP's of Flight Ops. and Industry Analysis should keep in mind that a "plan" is
simply a collection of workable ideas and a "critique" is just a systematic process of
asking "Why?" Your materials provide guidelines for finalizing your plan, summarizing
it, and critiquing the opposing plan.

3). When everyone agrees on a final plan, record the features of the plan in the Strategic
Planning Document form provided, sign it as a group, and notify the experimenter. Your
plan isn't valid and will not be implemented if it isn't signed by all members. Ifthis
happens, last year’s plan will be implemented again by default without any changes.

4). REMEMBER: YOU ONLY HAVE 75 MINUTES TO FINISH! When the timer
goes off, you must hand in what you have completed or your plan will not be valid.

173

APPENDIX D

APPENDIX D

Role Instructions Provided to the Vice-Presidents in the Dialectical Inquiry Process (1‘ DI
& SDI conditions).

Overview

Your role in the group discussion is similar to that of an attorney representing your Department
(i.e., Flight Operations or Industry Analysis). Your staff has put together some information that
will help improve next year's operations - your mission is to see that this plan is presented, its
information is considered during group discussion, and its best features make it into the final
plan. Your role is to summarize the essentials of the plan yorn' staff has put together, make sure
the group understands it, and "cross-examine" the other presenter to insure that he or she knows
what they're talking about.

Generating your plan:

Goal: Create a sound plan based on the information that you have at your disposal

1). REVIEW ALL YOUR INFORMATION and TAKE NOTES
2). LIST SPECIFIC CHANGES that will result in more profit
3). RECORD these changes on the last page of your yellow packet.

Presenting your plap: (iminmesonso)

Goal: Clearly and logically explain your plan so that others in the group know what it is
you want to do and why.

4). IDENTIFY problems with last year's operations

5). LIST YOUR SPECIFIC RECOMMENDATIONS for change
6). EXPLAIN your reasons

7). SUMMARIZE the advantages of your plan

Critigm_ng‘ the other le (5M0

Goal: Cross-examine the other presenter to discover their reasons, understand their plan
and ultimately identify weaknesses

8). FOR EACH MAJOR POINT, ASK "WHY?"

9). IDENTIFY PROBLEMS OR CONCERNS YOU HAVE WITH THEIR PLAN
10). EXPLAIN HOW YOUR PLAN AVOIDS THESE PROBLEMS

174

APPENDIX E

APPENDIX E

Instructions Provided to Groups in the Synthesis Dialectical Inquiry Condition

Your group will use the Competing Plans approach to help you create your strategic
plan. This simple approach is based on having two group members create their own individual
plans and present them to the group. The Vice Presidents of Flight Operations and Industry
Analysis have been randomly selected to create these plans. Note that your respective staffs have
already generated a number of good ideas. At the beginning of the Group Discussion Phase, the
two plans are presented and then each plan is critiqued by the other presenter. The presentations
should get a number of ideas out on the table for ftuther discussion, then the critiques will help to
"weed out" flawed ideas. Ideas and recommendations that survive the critique are more likely to
be good ideas than those that don't. Here is the sequence you should follow to implement the
Competing Plans approach:

At the start of th_e Group Discussion Phase:

1. The VP of Flight Operations presents Plan A (5 minutes or so)
2. The VP of Industry Analysis presents Plan B (5 minutes or so)
3. The VP of Flight Operations critiques Plan B (5 minutes or so)
4. The VP of Industry Analysis critiques Plan A (5 minutes or so)
5. Open Discussion (Remaining 55 minutes or so)

Hereare some gm'delines to follow in implementing the Commtipg Plans approach:

1). Again, the VP's of Flight Operations and Industry Analysis get to present and critique
plans. The VP's of Finance and Marketing should hold any questions and comments
until after the critiques. After the critiques have been conducted, everyone is free to say

anything.

2). The VP's of Flight Ops. and Industry Analysis should keep in mind that a "plan" is
simply a collection of workable ideas and a "critique" is just a systematic process of
asking "Why?" Your materials provide guidelines for finalizing your plan, summarizing
it, and critiquing the opposing plan.

 

3). The VP's of Marketing and Finance should listen carefully during the presentations and
critiques, take notes, and attempt to extract the best features of both plans. Afterwards,
these two Vice Presidents should summarize what has transpired, identify important
issues, and provide some structure for the remainder of the planning session. Your
materials provide guidelines for how to implement this facilitating role.

4). When everyone agrees on a final plan, record the features of the plan in the Strategic
Planning Document form provided, sign it as a group, and notify the experimenter. Your
plan isn't valid and will not be implemented if it isn't signed by all members. Ifthis
happens, last year's plan will be implemented again by default without any changes.

5). REMEMBER: YOU ONLY HAVE 75 MINUTES TO FINISH! When the timer
goes off, you must hand in what you have completed or your plan will not be valid.

175

APPENDIX F

APPENDIX F

Role Instructions Provided to Vice-Presidents in the Synthesis Role (SDI condition)

Role Instructions

Overview

Your role in the group is similar to that of a facilitator and discussion leader. During the
presentations and critiques, you should listen and try to understand what is being discussed.
When the critiques are finished, it will be up to you and the other facilitator to summarize, clarify
and focus the group so that you can come to some agreement.

During the debgg

Goal: Try to understand the essentials of the two plans that are presented by:

1). LISTENDIG carefully to both plans, noting any questions that you have.

2). TAKING NOTES

3). IDENTIFYING PROS AND CONS of each plan.

After the debate

Goal: Facilitate creation of a group plan by:

4). SUMMARIZING the advantages of each plan.

5). ASKING QUESTIONS and CLARIFYING points that are still confusing
6). SUGGESTING general approaches or specific actions

7). ORGANIZING topics for further discussion

176

APPENDIX G

APPENDIX G

Information Sharing Check-List Measure
Group #:

Coder:

 

Date Coded:

 

GROUP INFORMATION SHARING SHEET

Instructions to the Coder: Please indicate ____any piece of information that rs mken out loud
during the course of gzoap discussion by circlipg the rtem on th_e following sheets

 

 

 

Miscellaneoua (All): Miscellaawous (an Member Only):
1. Profit Formula 8. Route Fuel Costs Formula

2. Total Revenue Formula 9. Flight Staff Modifier Formula

3. Route Revenue Formula 10. Last Year’s Marketing Efions
4. Cash Formula 11. Marketing Coverage Rule

5. Costs Formula 12. Media Redundancy Formula

6. Debts Formula A 13. Expected Value Formula

7. Cmrent Assets 14. Expected Interest Rate

15. Ground Staff Formula
16. Flight Staff Formula
17. Retained % of Aircraft Sales

18. Probabilistic Natme of Competition

177

 

178

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

APPENDIX G
Lable 1. Flight Operations Data
Route Aircraft # Aircraft Flight Daily
Type Assigned Staff Flights
Memphis B-757 3 6 5
Louisville DC-9 3 6 6
Nashville DC-9 3 6 6
Raleigh-Durham A-300 4 4 7
Dallas-Ft. Worth DC-8 4 10 7
New Orleans B-757 3 7 5
Tampa L-101 1 3 8 6
Miami B-727 5 8 9
Chicago B-747 4 10 7
New York B-727 4 9 8
Los Angeles L-1011 3 12 6
Totals/Ave. - 39 - 72

 

 

179

 

 

 

 

 

 

 

 

 

 

 

 

 

 

APPENDIX G
Lable 2. Revenue Info_rrr_rati_or;
ROUTE PREVIOUS AVE. OUR PASS. MARKET
COMPETITION FARE FARE DEMAND SHARE

Memphis Low $250 $200 200,000 43%
Louisville Low $175 $250 125,000 35%
Nashville Moderate $200 $225 175,000 30%
Raleigh-Durham Moderate $175 $125 200,000 35%
Dallas- Heavy $275 $325 500,000 24%
Ft. Worth

New Orleans Heavy $250 $300 200,000 8%
Tampa Heavy $225 $250 225,000 16%
Miami V. Heavy $250 $300 525,000 8%
Chicago V. Heavy $300 $3 50 600,000 16%
New York V. Heavy $300 $300 550,000 15%
Los Angeles V. Heavy $400 $425 575,000 8%
Totals/Ave. — $254.55 $277.27 352,273 22%

 

 

 

 

 

 

 

 

Table 3. Cost Breakdown

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Type Cost % of Total
Aviation Fuel 65.91 M 36.6
Facilities/Equipment 48.00 M 26.6
Ground Staff 28.80 M 16.0
Flight Staff 25.6] M 14.2
Maintenance 5.85 M 3.2
Loan Repayment 2.75 M 1.5
Finance Charges 2.30 M 1.3
Advertising/Marketing 1.00 M 0.6
Total 180.22 M 100.0

 

180

APPENDIX G

Table 4. Profit Information by Route (in millions)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Route Route Route Profit Profit
Revenue Cost (millions) Ratio
Memphis $17.20 $10.36 $6.84 .66
Louisville $10.94 $1 1.75 - $ .81 -.07
Nashville $11.81 $10.96 $ .85 .08
Raleigh-Durham $ 8.75 $10.85 - $2.10 -.19
Dallas-Ft. Worth $39.00 $17.34 $21.66 1.25
New Orleans $ 4.80 $11.13 - $6.33 -.57
Tampa $ 9.00 $14.74 - $5.74 -.39
Miami $12.60 $20.09 - $7.49 -.37
Chicago $33.60 $19.77 $13.83 .70
New York $24.75 $20.54 $4.21 .20
Los Angeles $19.55 $32.69 - $13.14 -.40
Totals/Ave. $192.00 $180.22 $11.78 .08

 

 

Summary: Annual Fuel Costs by Route

181

APPENDIX G

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Route Rd. Trp. Cost/ Annual
Distance Flight Fuel Cost
Memphis 600 $900 $1.643 M
Louisville 600 $1080 $2.365 M
Nashville 400 $720 $1.577 M
Raleigh-Durham 500 $500 $1.278 M
Dallas-Fort Worth 1000 $2300 $5.87 7 M
New Orleans 800 $1200 $2.190 M
Tampa 800 $2200 $4.818 M
Miami 1100 $2310 $7.588 M
Chicago 1300 $3250 $8.304 M
New York 1400 $2940 $8.585 M
Los Angeles 3600 $9900 $21.68] M
Totals/Ave. 1100 $2445 $65.91 M
Smary: Aircraft Characteristics & Cost Information
Type Accom. Flight Staff Cost/Mile Purchase
Class Lim. Cost (mil).
A-300 E 4 $1.00 $1.0 M
DC-9 8 $1.80 $2.0 M
B-757 D 7 $1.50 $2.5 M
L-1011 C 12 $2.75 $3.0 M
B-727 C 9 $2.10 $4.5 M
DC-8 B 11 $2.30 $6.0 M
B-747 B 15 $2.50 $6.0 M
DC-10 A 12 $2.40 $8.0 M

 

 

 

182

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

APPENDIX G
Modifier #1: Convenience (Daily Flights)
COMPETITION
Daily Low- Heavy-
Flights Moderate Very Heavy
1 -12% -25%
2 -10% -16%
3 -8% -11%
4 -5% -7%
5 -2% -4%
6 0% -2%
7 +2% 0%
8 +4% 0%
9 +6% +1%
10 +8% +3%
11 +9% +5%
12 +10% +7%
13 +11% +8%
14 +11% +9%

 

 

 

 

 

Modifier #2: Comfort (Accommodations)

 

 

 

 

 

 

 

Roundtrip Distance

Factors <1500 miles 1501+ miles

Competition Low or Mod. Heavy or Low or Mod. Heavy or
Accommodations V- Heavy V- Heavy
"E" Class -2% -10% -20% -35%
"D" Class 0% -5% -10% -20%
"C" Class 0% 0% -l% -7%
"B" Class +2% +5% +7% +5%
"A" Class +3% +7% +12% +9%

 

 

 

 

 

 

 

Previous and Projected Costs

183

APPENDIX C

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Item Last Year Projected Next Yr.
Aviation Fuel (tot.) $65,910,000 ——
Ground Staff personnel @ $40,000 $42,000
Flight Staff personnel @ $45 ,000 $49,000
Maintain Domestic Facility @ $4,000,000 $4,250,000
Start-up Domestic Facility @ -— $5 ,000,000
Start-up Foreign Facility @ -- $7,000,000
Aircraft Maintenance @ $150,000 $160,000
Marketing (tot.) $ 1,000,000 ---
Loan Repayment (A) $ 1,500,000 $ 1,500,000
Loan Repayment (B) $1 ,250,000 $1 ,250,000
Finance Charges (A) $633,750 $536,250
Finance Charges (B) $1,663,750 $1,526,250
Loan I.D. Outstanding Interest Rate

Principal
001-91 (A) 9,000,000 6.5%
002-94 (B) 14,500,000 11.0%

 

 

 

 

 

184

APPENDIX G

Relationship between Lare Price and Market Shar_e.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Competition
Deviation from Low Moderate Heavy Very Heavy
Average Fare
$ 100 under ave. 53% 44% 35% 26%
$75 under ave. 48% 39% 31% 25%
$50 under ave. 44% 35% 28% 24%
$25 under ave. 41% 32% 25% 21%
Same as Ave. 40% 30% 22% 17%
$25 over ave. 39% 28% 19% 13%
$50 over ave. 37% 24% 16% 10%
$75 over ave. 34% 20% 12% 9%
$100 over ave. 30% 15% 9% 8%
Marketing Modifiers
Media Cost/City Effect/Route - Effect/Route -
(Low/Moderate (Heavy/V. Heavy
Competition) Competition)
Television $5,000,000 +3-4% +2-5%
Radio $3,000,000 +2-3% +2-3%
Newspaper $2,000,000 +2% +l-3%
Billboard $1,000,000 +1% +l-2%
NONE O 0% -3%

 

 

 

 

 

 

185

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

APPENDIX G
Existing Routes: ATLANTA—
Route Demand Rd. Trp. Competition Ave.
Distance Likelihood Flight Staff
Memphis 195-205,000 600 70-15-15-0 6
Louisville 100-1 50,000 600 75-15-10-0 6
Nashville 170-210,000 400 60-30-10-0 5
Raleigh- 200-225,000 500 50-35-15-0 5
Durham
Dallas- 425-475,000 1,000 5-20-60-15 9
Ft. Worth
New Orleans 190-220,000 800 15-35-35-15 7
Tampa 215-245,000 800 5-30-50-15 9
Miami 500-530,000 1,100 0-15-65-30 10
Chicago 605-625,000 1,300 0-5-15-80 10
New York 560-580,000 1,400 0-10—15-75 1 1
Los Angeles 585-610,000 3,600 0-10-20-70 12
Potential Routes: ATLANTA-
Route Demand Rd. Trp. Competition Ave. Ave.
Distance Likelihood Fare Flight Staff
Madrid 85-90,000 8,000 70-15-15-0 $875 12
Paris 75-125,000 9,000 55-20-20-5 $900 1 1
London ISO-200,000 9,000 50-30-10-10 $850 12
Sao Paulo-Rio 80-100,000 12,000 85-15-0-0 $1200 10
Mexico City 110-175,000 2,800 40-30-20-10 $500 11
Cancun 90-120,000 3,000 55-30-10-5 $500 13
Virgin Islands 85-145,000 4,000 45-35-15-5 $550 11

 

 

 

 

 

 

 

 

186

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

APPENDIX C
Potential Routes: ATLANTA-
Route Demand Rd. Trp. Competition Ave. A. Flight
Distance Likelihood Fare Staff
Seattle 305-350,000 4,500 50-30-15-5 450 12
Minneapolis- 255-295,000 1,800 65-30-5—0 325 l 1
St. Paul
Cincinnati 265-295,000 700 60-20-15-5 250 7
Indianapolis 245-260,000 800 75-20-5-0 225 9
San Francisco 395-470,000 4,200 40-20-20-20 425 11
Denver 420-500,000 2,300 20-40-30-10 375 12
St. Louis 260-280.000 900 25-60-10-5 250 6
Buffalo 270-295,000 1,400 35-30-30-5 300 8
Kansas City 245-255,000 1,300 30-55-15-0 300 9
Pittsburgh 385-450,000 1,000 10-25-55-10 275 8
Philadelphia 335-365,000 1,200 15-25-50-10 300 7
Phoenix 395-490,000 3,100 10-65-20-5 375 13
Boston 475-520,000 1,800 10-15-30-45 325 9
Detroit SOD-615,000 1,200 0-10-25-65 325 8
Wzshington, 725-800,000 1,000 0-10-20-70 250 9
D. .

 

 

 

 

 

 

 

Miscellfleous Other:

 

APPENDIX H

APPENDIX H

Observer Rating Form: Intragroup Conflict

0 = Can't recall any instances

1 = A few minor, isolated instances

2 = Several major, isolated instances

3 = Ongoing occurrences throughout discussion

Please indicate the extent to which you observed instances of the following during group
interaction/discussion:

1. Negative Affect

-Long, awkward Silence

-Frustration

-Physica1 agitation (fidgeting, squirming, tapping, etc.)
-Annoyed tone of voice

-Evident Tension

2. Hostile/Degrading Interpersonal Behavior

-Mocking/sarcastic tone of voice
—Condescension/"lect1ning" other members
-Insults/Snide Remarks

-Efforts to force others to agree or not agree

3. Irrational Thought Processes

—Stubbom refusal to acknowledge appropriateness of reasonable
explanations
-Meandering conversations unrelated to real issues the group has
identified
-Settling disagreements by arbitrary means, or compromise plans that are
clearly political rather than based on the merits of the ideas.

4. TOTAL (SUM of #s 1.3)

187

APPENDIX I

APPENDIX I

Questionnaire Items: Intragroup Conflict

Response Scale:

1 = Strongly Disagree

2 = Disagree

3 = Sort of Disagree, Sort of Agree
4 = Agree

5 = Strongly Agree

11.

2.

3.**

8.**

10.

There was a good deal of tension in the group during the simulation.
There was no hostility in the group during this task.

At times, voices were raised during discussion.

At times the group didn't seem to care if we came up with a good plan.
Sometimes I felt the group couldn't comprehend simple logic.

There was a lot of fi'ustration in our group during this study.

People got angry dming planning.

We got along well as a group.

There were times when the group considered ideas that were clearly
unreasonable.

There was a lot of arguing during the group planning phase.

"Indicates items removed from original scale as a result of poor item-total correlations

188

APPENDIX J

APPENDIX J

Observer Rating Form: Process Facilitation

0 = No instances recalled

I = A few isolated instances recalled

2 = A moderate number of instances recalled
3 = Ongoing throughout discussion

Please indicate the extent to which you observed the following during group
interaction/discussion:

1. Reflecting/Summarizing:
Statements that paraphrase or restate what another group member has just said in
one form or another for the purpose of ensuring that communication was clear
and that the listener(s) understood the speaker correctly. Consider things such as
paraphrasing or highlighting what OTHER members have said, or repeating back
ideas, suggestions or recommendations to insure the listener or group understood
them correctly.

2. Clarifying:
Statements addressed to a group member or members that seek to clarify
information, problems or issues faced by the entire group so that the entire group
can work from a common perspective. Consider things like attempts to sort out
confusing ideas expressed by other group members, asking for information that
is needed to make sense of an issue, "framing" an issue or defining a problem
that has the group hung up.

3. Integrating:
Assertive statements that attempt to draw together two or more distinct lines of
thought, and/or attempts to consolidate move the discussion forward. Consider
things like attempts to combine the best ideas presented by two or more group
members into an integrated plan

4. Focusing/Structuring:
Statements that move the group forward by identifying areas where the group has
reached agreement, and statements that identify areas which the group still needs
to consider. Consider things like announcements similar to "I think we all agree
that. . .," suggesting the next topic for discussion, proposing a timetable, etc.

TOTAL (SUM of #s 1.4)

Note: Only consider things that are noticeable enough that all or most other group members
would have perceived them.

189

APPENDIX K

APPENDIX K

Observer Rating Form: Controversy

0 = Can't recall any instances

1 = A few minor, isolated instances

2 = Several major instances

3 = Ongoing occurrences throughout discussion

Please indicate the extent to which you observed instances of the following during group
interaction/discussion:

1. Indirect Challenge
Questions directed to a group member that imply that the speaker doesn't agree
with a suggestion or recommendation, or doesn't understand reasons that have
been offered. Includes things like presenting contradictory information in
response to a suggestion or recommendation, asking "Why?," or agreeing with
only part of what somebody else says.

2. Explicit Disagreement
DIRECT statements that show the speaker does not agree with a comment, idea
or recommendation, or the information that supports it. Includes things like
saying "I don't agree," "I think you're wrong," or "Not according to my
information." [Do not include hints or subtle suggestions]

3. Presentation of Opposing Viewpoints
Presentation of a set of specific recommendations, a general approach or a
central idea that is incompatible with what has been suggested by one or more
other group members. [Do not consider minor comments made in passing -
must be a deliberate effort attended to by the rest of the group.]

4. TOTAL (SUM of #s 1-3)

Only consider comments or behaviors that are noticeable enough that all or most other group
members would have perceived them.

190

APPENDIX L

APPENDIX L

Observer Rating Form: Controversy (Recoded)

0 = None

1 = Superficial (Mention B)

2 = Moderate (Support for A [or B] from someone)
3 = Severe (Support for B [or A] from someone)

4 = Beyond Level 3

Controversy is defined as real or apparent disagreement within the group concerning task-
specific strategies the group should use and/or how the group should structure its activities as it
completes the task. Controversy is impersonal; it refers to disagreement over ideas. Do not
confuse this with interpersonal conflict! Please indicate the extent to which you observed
DISAGREEMENT over the following things during group interaction/discussion.

Task-Related Controversy: Issues and Strategy

1.

2.

Expansion v. Consolidation

High Fare v. High Market Share

Fuel Efficiency v. Accommodations

Increasing Revenue v. Cutting Costs

Where/How Should Cash Be Spent (Marketing, New Aircraft, Loans, Invest)

Free-Floating (Cannot be categorized)

Process-Related Controversifiz Activities and Decision Making

7.

How Group Should Identify Ideas

8. How Group Should Approach Task (Serial v. Whole)

9. Who Should Make Decisions (Majority Rule v. Expert)

10. TOTAL (SUM of #s 1-9)

191

APPENDIX M

APPENDIX M

Task Knowledge Measures
GENERAL

The purpose of this check is to help you assess the degree to which you understand the
information you have been provided. Please find the answers to the following questions using
any and all materials provided to you. Feel free to change your answers until you are satisfied
you have them right. If you can answer all or most of the questions, you will be a good
representative for your department. PLEASE COMPLETE THIS FORM BY THE END OF
GROUP DISCUSSION.

General Concepts: Please put your answer below the question.

1. All flights must be non-stop between some city and where?
2. How many roundtrip flights can ONE aircraft make in a single day?
3. How much money does your firm currently have in CASH?

4. How many B-727's does your firm currently own?

5. Route Revenue increases as three other variables increase. What are these three
variables?
a.

b.
c.
6. Which route offered the most Daily Flights last year?
7. Which route earned the highest Market Share last year?
8. Which route had the highest Passenger Demand last year?

9. Which route lost the most money last year?

10. Which route had the worst profit ratio last year?

192

APPENDIX M

FLIGHT OPERATIONS

1. What is the DIFFERENCE in Market Share between having 3 Daily Flights on a route
and 9 Daily Flights when there is Heavy Competition?

2. What is the modifier to Market Share for having aircraft with "D" class Accommodations
on a route with Roundtrip Distance = 2000 miles and Moderate Competition?

3. What is the modifier to Market Share on a route with Very Heavy Competition and 3
more Flight Staff on each route than the industry average?

4. How much would it cost per year in aviation fuel for a route where:
a. Cost/Mile of the aircraft used = $1.90
b. Roundtrip distance = 1,000 miles
c. Daily Flights = 10

5- What would be the cost SAVINGS for the above route if the Cost/Mile was 1.25 and
there were only 5 Daily Flights?
(Remember to find the difference.)

194

APPENDIX M
FINANCE
1. How much money can you expect to make on the sale Of an aircraft listed at $6,000,000?
2. If your firm offered 100 Total Daily Flights on all routes together, how many Ground
Staff would be required?

3. According to your staff‘s estimates, what would be the cost of paying 750 Ground Staff
next year?

4. How many TOTAL Flight Staff personnel are needed for a route with 7 Flight/Staff per
flight and 6 Daily Flights?

5- How much EXTRA does your staff estimate it would cost to Start-Up a Foreign Facility
next year as compared to Maintaining an Existing Domestic Facility?
(Remember to find the difference.)

 

195

APPENDIX M

MARKETING

1. What is the Market Share you would earn if you were charging $75 LESS THAN the
average on a route with Moderate Competition?

2. What is the estimated bonus effect on Market Share for a route with Low Competition
where TV, radio, newspaper and billboard advertising are ALL used (assuming
maximum possible benefit for each medium)?

3. What is the modifier to Market Share for routes with Heavy or Very Heavy Competition
if you don't use any form of Advertising?

4. What is the expected value for a route where:
Price of Fare = $300
Average Fare = $250
Competition = Very Heavy

5- Using the Passenger Revenue formula, how much EXTRA PROFIT would you make by
charging $2 50 as opposed to $3 00 in the previous question, assuming Passenger Demand
= 500,000 for the year? (Remember to find the difference.)

196

APPENDIX M

INDUSTRY ANALYSIS

3-5.

Of the cities NOT currently serviced by SouthEast (i.e., potential cities), which cities will
definitely have a Passenger Demand greater than 400,000 next year?

Which POTENTIAL routes have a 50% (or greater) chance of having Low Competition
next year?

Assuming the maximum possible Passenger Demand and the most likely Competition
Level (based on probabilities given), which three potential cities would bring in the most
revenue next year?

APPENDIX N

APPENDD( N

Questionnaire Items: Implementation Quality

1. Please identify the inquiry method your group used during this study?

 

2. Please identify your role within the group

 

Please use the following scale to respond to the next four items:

1 = Strongly Disagree

2 = Disagree
3 = Undecided
4 = Agree

5 = Strongly Agree

1. I feel that I did a good job performing my role in the group.
2. Most of the time, I wasn't conscious of the role I was supposed to be playing.

3. Most of the other members of my group performed their role adequately.

197

LIST OF REFERENCES

LIST OF REFERENCES

Ackofl, R.L. (1974). Redesim’ g the future. New York: Wiley.

Aldag, R.J., & Fuller, SR. (1994). Beyond fiasco: A reappraisal of the groupthink
phenomenon and a new model of group decision processes. Psychological
Bulletin. 113. 533-552.

Argote, L., & McGrath, J .E. (1993). Group processes in organizations: Continuity and
change. In C.L. Cooper and LT. Robertson, (Eds), International Review of
Organizational Psychology, 8, 333-389.

Bettenhausen, KL. (1991). Five years of groups research: What we have learned and
what needs to be addressed. Journal of Management, _1_7, 345-381.

Brehmer, B. (1976). Social judgment theory and the analysis of interpersonal conflict.
Psychological Bulletin, 8_3, 985-1003.

Callaway, M.R., & Esser, J .K. (1984). GrOupthink: Effects of cohesiveness and
problem-solving procedures on group decision making. Social Behflor a_na
Personalig, 12, 157-164.

Cartwright, C., & Zander, A. (Eds). (1968). Group dmamics: Research and theory.
New York: Harper & Row.

Chanin, M.N., & Shapiro, H]. (1984). Dialectical and Devil's Advocate Problem-
Solving. Asia Pacific Journal of Management, 1, 159-170.

Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2ad ed.).
Hillsdale, NJ: Erlbaum.

Cook, R.L., & Hammond, KR. (1982). Interpersonal learning and interpersonal conflict
reduction in decision-making groups. In R. Guzzo (Ed.), Improving group

decision making in organizations: Approaches from theogy and research (pp. 13-
40). New York: Academic Press.

198

199

Cosier, R. (1978). The effects of three potential aids for making strategic decisions on
prediction accuracy. Qrganizational Behavior and Human Performarae, 22, 295-
306.

Cosier, R. (1980). Inquiry method, goal difficulty, and context effects on performance.
Decision Sciences, _1_1_, 1-16.

Cosier, R., & Aplin, J .C. (1980). A critical view of dialectical inquiry as a tool in
strategic planning. Strategic Management Journal, 1, 343-356.

Cosier, R., & Rechner, PL. (1985). Inquiry method effects on performance in a
simulated business environment. Organizational Behavior and Human Decision
Processes, 3g, 79-95.

Cosier, R., & Rose, G.L. (1977). Cognitive conflict and goal conflict effects on task
performance. Organizational Behavior and Human Performance fl, 378-391.

 

Cosier, R., Ruble, T.A., & Aplin, J.C. (1978). An evaluation of the effectiveness of
dialectical inquiry systems. Management Science, _2__4_, 1483-1490.

Courtright, J .A. (1978). A laboratory investigation of groupthink. Communications
Monogrgaphs, 5, 229-246.

Devine, D. (1995). SouthEast Airlines, Inc.: An organizational simulation.

Emshoff, J .R., & Finnel, A. (1978). Defining corporate strategy: A case study using
strategic assumptions analysis. Wharton Applied Research Center. working paper
no. 8-78.

Esser, J .K., & Lindoerfer, J .S. (1989). Groupthink and the space shuttle Challenger
disaster: Toward a quantitative case analysis. J omnal of Behavioral Decision
Making, 2, 167-177.

F ahey, L. (1981). On strategic management decision processes. Strategic Management
Journal, 2, 43-60.

F lowers, ML. (1977). A laboratory test of some implications of Janis' groupthink
hypothesis. Journal of Personality and Social Psychology, fl, 888-896.

Fodor, E.M., & Smith, T. (1982). The power motive as an influence on group decision
making. Journal of Personality and Social Psycholagy, _4_2_, 178-185.

Frost, J.H., & Wilmot, W.W. (1978). Interpersonala conflict. Dubuque, IA:
William C. Brown & Co.

 

200

Futrell, D.A., & Sundstrom, E. (1993). Group composition and performance: Cognitive
ability and group productivity in an assembly task. Paper presented at the annual
meeting of the Society for Industrial and Organizational Psychology, San
Francisco.

Galbraith, JR. (1973). Desigm'hg complex organizations. Reading, MA: Addison-
Wesley

George, A. (1972). The case for multiple advocacy in making foreign policy. American
Political Science Review, _6_6, 751-785.

 

 

f”

Hackman, J .R. (1988). The design of work teams. In J .W. Lorsch (Ed.), Handbook of
organizational behavior (pp. 315-342). Englewood Cliffs, NJ: Prentice-Hall. ,
L-fi

Hackman, J .R. (Ed.). (1990). Groaps that work (and those that don't): Creating
conditioaa for effective teaflwork. San Francisco: J ossey-Bass.

Hackman, J.R., & Morris, CG. (1975). Group tasks, group interaction processes, and
group performance effectiveness: A review and proposed integration. In L.
Berkowitz (Ed.), Advanaces in experimental social psychology (V 01. 8, pp. 45-99).
San Diego: Academic Press.

Hammond K.R., Todd, F .J ., Wilkins, M., & Mitchell, TD. (1966). Cognitive conflict
between persons: Applications of the "lens model" paradigm. Journal of
Exgrimental Socail Psychology,2, 343-360.

Hensley, T.R., & Griffin, G.W. (1986). Victims of groupthink: The Kent State
University Board of Trustees and the 1977 gymnasium controversy. Journal of
Conflict Resolutioh, 3_Q, 497-531.

Herbert, T.T., & Estes, R.W. (1977). Improving executive planning by forrnalizing
dissent: The corporate devil's advocate. Academy of Management Review, _2_,
662-667.

Hill, G.W. (1982). Group versus individual performance: Are n+1 heads better than
one? Psychological Bulletin, 21, 517-539.

Hirokawa, R.Y. (1990). The role of communication in group decision making efiicacy:
A task-contingency perspective. Small Group Research, _2_1, 190-204.

Ilgen, D.R., Major, D.A., Hollenbeck, J .R., & Sego, DJ. (1993). Team research in the
1990s. In M.M. Chemers and R. Ayaman (Eds), Leadership theory and reseaach
(pp. 245-270). New York: Academic Press.

201

Ilgen, D.R., Major, D.A., Hollenbeck, J .R., & Sego, DJ. (1993). Team research in the
19905. In M.M. Chemers and R. Ayaman (Eds), Leadership theoryand research
(pp. 245-270). New York: Academic Press.

James, LR. (1982). Aggregation bias in estimates of perceptual agreement. Journal
of Applied Psychology, 6_7, 219-229.

Janis, LL. (1972). Victims of groupthink. Boston: Houghton Mifllin.

Janis, IL. (1982). Gr_p__ou think: Psycholggjcal studies of policy decisions and fiascos.
Boston: Houghton Mimin.

Janis, I.L., & Mann, L. (1977). Decision Making: A psychological analysis of conflict.
choice and commitment. New York: Free Press.

Katz, D., & Kahn, R.L. (1976). The social psychology of organizations (2nd ed.). New
York: Wiley.

Koopman, P.L., & Pool, J. (1990). Decision making in organizations. International
Review of Industrial and Organizational Psychology, 5, 101-148

Larson, J. R. & Christensen, C. (1993). Groups as problem-solving units. Towards a
new meaning of social cognition. British Journal of Social Psychology,__, 32 5- 30.

Latane, B. (1986). Responsibility and effort in organizations. In P.S. Goodman (Ed.),
m effective work groups. (pp. 277-304). San Francisco: Jossey-Bass.

Latane, B., Williams, K., & Harkins, S. (1979). Many hands make light the work: The
causes and consequences of social loafing. J ourryal of Person_ality and Social
Psychology, 3_7_, 822-832.

Laurenco, S.V., & Glidewell, J.C. (1975). A dialectical analysis of organizational
conflict. Administrative Science Quarterly, 29, 489-508.

Leana, CR. (1985) A partial test of Janis' groupthink model: Effects of group
cohesiveness and leader behavior on defective decision making. Journal of

Management, fl, 5-17.

Levine, J .M., & Moreland, R.L. (1990). Progress in small group research. Annual
geview of Psychology, fl, 585-634.

Lorge, I., & Solomon, H. (1955). Two models of group behavior in the solution of
eureka-type problems. byehometrilg, Q, 139-148.

202

Marquart, DJ. (1955). Group problem solving. Journal of Social Psycholggy, 4_1, 103-
113.

Martin, R.J. (1983). A skills and strategies handboojk for working with people.
Englewood Cliffs, NJ: Prentice-Hall.

Mason, RD. (1969). A dialectical approach to strategic planning. Management Science,
Q, 404-414.

Mason, R.D., & Mitrofi, 1.1. (1981). Challengm' g strategic planning assumptions. New
York: Wiley-Interscience.

McGrath, J .E. (1984). Groups: Interaction and performance. Englewood Clifl‘s, NJ:
Prentice-Hall.

McGrath, J .E., & Altman, I. (1966). Small ggogp research: A synthesis and
cringe of the field. New York: Holt, Rinehart & Winston.

Milliken, F.J., & Martins, LL. (1996). Searching for common threads: Understanding
the multiple effects of diversity in organizational groups. Academy of
Management Review. A, 402-433..

Mintzberg, H., Raisinghani, D., & Theoret, A. (1976). The structure of 'unstructured'
decision processes. Administrative Science Quarterly. 2_1_, 246-275.

Mitroff, 1.1. (1982). Dialectic squared. Decision Sciences. 13, 222-224.

Mitroff, 1.1., Barabba, V.P., & Kilmann, RH. (1977). The application of behavioral and
philosophical technologies to strategic planning: A case study of a large federal
agency. Management Science. 2_4, 44-58.

Mitrofl‘, I.I., & Emshoff, J .R. (1979). On strategic assumption-making: A dialectical
approach to policy and planning. Academy of Management Review. 4, 1-12.

Mitrofl‘, I.I., Emshofl‘, J.R., & Kilmann, RH. (1979). Assumptional analysis: A
methodology for strategic problem solving. Management Science. 25, 583-593.

Mitroff, I., & Mason, R. (1981). The metaphysics of policy and planning: A reply to
Cosier. Academy of Management Review. 9, 649-652.

Moorhead, G. (1982). Groupthink: Hypothesis in need of testing. Group and
Organization Studies, 1, 429-444.

203

Moreland, R.L., & Levine, J.M. (1992). The composition of small groups. In L.
Berkowitz (Ed.), Advances in Group Processes (V 01. 9, pp. 237-280). New York:
JAI Press.

Mullen, B., & Copper, C. (1994). The relation between group cohesiveness and
performance: An integration. Psychological Bulletin,.

Patton, B.R., Griffin, K., & Patton, EN. (1989). Decision making and goup
interaction (third edition). New York: Harper Collins.

Pearce, J .A., H, & Robinson, RB., Jr., (1988). Strategic management: Strategy
formul_a£ion and implementation (3rd edition). Homewood, IL: Irwin.

Posner-Webber, C. (1987). Update on groupthink. Sm“ Group Behavior. 3, 118-125.

Priem, R., & Price, K.H. (1991). Process and outcome expectations for the dialectical
inquiry, devil's advocacy, and consensus techniques of strategic decision making.
Gropp and Organizaiion Studies, _1_6, 206-225.

Pruitt, D.G., & Rubin, J.Z. (1986). Social conflict: Escalation. stalemate and
settlement. New York: Random House.

 

Quinn, J.B. (1980). Strategies for change: Logical incremergalism. Homewood: Irwin.
Rahim, M.A. (1986). Manag1p' g conflict in organization; New York: Praeger.

Roberts, K.H., Hulin, C.L., & Rousseau, D.M. (1978). Developing an
interdisciplinary science of organizations. San Francisco: J ossey-Bass.

Rousseau, D. (1985). Issues of level in organizational research: Multilevel and cross-
level perspectives. In L.L. Cummings & B.M. Staw (Eds), Research in
Organizational Behavior, (Vol. 7, pp. 1-37). Greenwich, CT: JAI Press.

Salas, E., Dickinson, T.L., Converse, S.A., & Tannenbaum, S. (1992). Toward an
understanding of team performance and training. In R.W. Sweezey & E. Salas

(Eds), Teams: Their training and mrformance. Norwood, NJ: Ablex.

Saveedra, R., Barley, P.C., & Van Dyne, L. (1993). Complex interdependence in task-
performing groups. Journal of Applied Paychology, 18, 61-72.

Schweiger, D., & Finger, P. (1984). The comparative effectiveness of dialectical inquiry
and devil's advocacy: The impact of task biases on previous reseal of Behavioral
Decision Making, 3, 229-245.

204

Schweiger, D., Sandberg, W., & Ragan, J. (1986). Group approaches for improving
strategic decision making: A comparative analysis of dialectical inquiry, devil's
advocacy, and consensus. Agdemy of Management Journal, 2, 51-71.

Schweiger, D.M., Sandberg, W.R., & Rechner, PL. (1989). Experiential effects of
dialectical inquiry, devil's advocacy, and consensus approaches to strategic
decision making. Academy of Management Journal, 3_2, 745-772.

Schwenk, CR. (1982). Effects of inquiry methods and ambiguity tolerance on prediction
performance. Decision Sciences, 1, 841-855.

Schwenk, OR. (1984). Inquiry method effects on prediction performance: Task
involvement as a mediating variable. Decision Sciencea, Q, 449-462.

Schwenk, CR. (1985). Effects of planning aids and presentation media on
performance and affective responses in strategic decision making. Management
Science, 39, 263-272.

Schwenk, CR. (1988). The essence of strat_egic decision making. Lexington, MA: D.C.
Heath Co.

Schwenk, OR. (1990). Effects of devil's advocacy and dialectical inquiry on decision
making: A meta-analysis. Organizational Behaviflnd Human Decision
Processes fl, 161-176.

 

Schwenk, C.R., & Cosier, R.A. (1980). Efi‘ects of expert, devil's advocate, and
dialectical inquiry methods on prediction performance. Organizational Behavior
and Human Performance, E, 409-424.

Schwenk, C.R., & Cosier, R.A. (1993). Effects of consensus and devil's advocacy on
strategic decision making. Journal of Applied Social Psychology, Q, 126-139.

Schwenk, C.R., & Thomas, H. (1983). Efl‘ects of conflicting analyses on managerial
decision making. Decision Sciences. 11, 467-482.

Shaw, M. (1932). A comparison of individuals and small groups in the rational solution
of complex problems. American Journal of Paycholgy, fl, 491-504.

Shaw, ME. (1981). Ma: The psychology of small gioup behavior (3rd
Ed.). New York: McGraw-Hill.

 

Shirley, RC. (1982). Limiting the scope of strategy: A decision based approach.
Academy of Management Review, 1, 262-268.

205

Smith, S. (1984). Groupthink and the hostage rescue mission. British Journal of
Political Science, 1_5, 117-126.

Smith, K.K., & Berg, Organizational Behavior & Human Decision Processes. 41, 161-
176.

Smith, K.K., & Berg, D.N. (1987). A paradoxical conception of group dynamics.
Human Relations. fl, 633-657.

Stanley, JD. (1981). Dissent in organizations. fiademy of Management Reyiew,
a, 13-19.

Stasser, G., & Stewart, D. (1992). Discovery of hidden profiles by decision making
groups: Solving a problem versus making a judgment. Journal of Personality aaci
Social Psychology, 6_3, 426-434.

 

Stasser, G, Taylor, L.A., & Hanna, C. (1989). Information sampling in structured
discussions of three- and six-person groups. Journal of Personality and Social
Psychology, 51, 67-78.

Stasser, G., & Titus, W. (1985). Pooling of unshared information in group decision
making: Biased information sampling during discussion. Journal of Personaligy
and Social Psychology, fl, 1467-1478.

Stasser, G., & Titus, W. (1987). Effects of information load and percentage of shared
information on the dissemination of rmshared information during group
discussion. Journal of Personality m Social Psychology, fl, 81-93.

Steiner, ID. (1972). Group process and productivity. Orlando, FL: Academic Press.

Taylor, RN. (1992). Strategic decision making. In M.D. Dunette and L.M. Hough

(Eds), Handbook of Industrial and Organizational Psychology (Vol. II, 2nd
edition). Palo Alto, CA: Consulting Psychologists Press.

Tetlock, RE. (1979). Identifying victims of groupthink from public statements of
decision makers. Journal of Personality and Social Psychology, 3;, 1314-1324.

Tetlock, P.E., Peterson, R.S., McGuire, C., Change, S., & Feld, P. (1992). Assessing

political group dynamics: A test of the groupthink model. Journal of Personaliiy
and Social Psychology, 6_3, 403-425.

Thomas, H. (1984). Strategic decision analysis: Applied decision analysis and its role in
the strategic management process. Strategic Management Journal, Q, 139-156.

206

Tjosvold, D. (1985). Managerial implications of controversy research. Journal of
Mmgement, fl, 221-238.

 

Tjosvold, D. (1986). Working together to get things done: Managipg for organizational
productivity. Lexington, MA: Lexington Books.

Tziner, A., & Eden, D. (1985). Effects of crew composition on crew performance:
Does the whole equal the sum of the parts? J oumai of Applied Psychology, 10,
85-93.

Wall, V.D., Galanes, G.J., & Love, SB. (1987). Small, task-oriented groups: Conflict,
conflict management, satisfaction and decision quality. Small Group Behavior,
_1_8, 31-55.

Walsh, J .P., & Fahey, L. (1986). The role of negotiated belief structures in strategy
making. Journal of Management, Q, 325-338.

Walsh, J.P., Henderson, C.M., & Deighton, J. (1988). Negotiated Belief Structures and
Decision Performance: An empirical investigation. Organizational Behavior &
Hurpan Decision Processes, 32, 194-216.

Whyte, G. (1989). Groupthink reconsidered. Academy of Management Review, 15, 40-
56.

Wood, W. (1985). Meta-analytic review of sex differences in group performance.
Psychological Bulletin. 102. 53-71.

 

"71111111111111.11111111111“