Foundations For Group Decision Analysis

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/262176101
Foundations for Group Decision Analysis
Article in Decision Analysis · June 2013

DOI: 10.1287/deca.2013.0265
CITATIONS READS
67 1,527
1 author:
Ralph L. Keeney
Duke University
234 PUBLICATIONS 31,464 CITATIONS
SEE PROFILE
All content following this page was uploaded by Ralph L. Keeney on 13 September 2015.
The user has requested enhancement of the downloaded file.

Decision Analysis
Articles in Advance, pp. 1–18
posted on any other website, including the author’s site. Please send any questions regarding this policy to permissions@informs.org.
Copyright: INFORMS holds copyright to this Articles in Advance version, which is made available to subscribers. The ﬁle may not be
ISSN 1545-8490 (print) ISSN 1545-8504 (online) http://dx.doi.org/10.1287/deca.2013.0265

© 2013 INFORMS
Foundations for Group Decision Analysis

Ralph L. Keeney
Fuqua School of Business, Duke University, Durham, North Carolina 27708, keeney@duke.edu
T his paper derives a general prescriptive model for group decision analysis based on a set of logical and
operational assumptions analogous to those for individual decision analysis. The approach accounts for each
group member’s potentially different frames of their common decision, including different events and different
consequences of concern. Assuming that each group member accepts the decision analysis assumptions to
evaluate his or her analysis of what the group should do and that the group accepts an analogous set of decision
analysis assumptions for the group’s decision, it is proven that the group expected utility for an alternative
should be a weighted sum of the individual member’s expected utilities for the alternatives. After each group
member does his or her decision analysis of the group’s alternatives, the essence of the group decision analysis
is to specify the weights based on the interpersonal comparison of utilities and on the relative importance or
power of each individual in the group.
Key words: group decision analysis; expected utility; group decisions; decision analysis
History: Received on December 20, 2012. Accepted by former Editor-in-Chief L. Robin Keller on January 6,
2013. Published online in Articles in Advance.
1. Introduction rule, such as the candidate with the most votes wins.
Decision analysis is a methodology and set of proce- In this case, there is a collection of individual deci-
dures for building models to guide decision making. sions by the voters that leads, with no specific group
These models are prescriptive in that they provide action, to a selected alternative. The third class of
insights about what alternative to choose to best decisions is social planning or social welfare deci-
achieve stated objectives. The foundations for decisions where an individual planner or organization,
sion analysis, which are a set of logical and oper- after taking judgments and preferences of individuals
ational assumptions (Pratt et al. 1964), are well affected by the decision into account, makes the deci-
developed for individual decisions. However, several sion. The model developed here may have relevance
attempts to extend the appealing logic of decision to each of these classes of decisions, but such uses are
analysis to group decisions have not succeeded except not the topic of this paper.
in special cases. This paper develops a general group To provide prescriptive guidance for group deci-
decision analysis model. sions, the inherent reasons for their complexity must
Group decisions, as defined in this paper, are deci- explicitly be addressed in the group decision analysis.
sions where a group of two or more individuals must These reasons include the following:
collectively select an alternative from a set of two 1. The members of the decision-making group may
or more alternatives that best satisfies the group’s view their common decision differently. They may be
objectives, and no individual has veto power. This concerned about different consequences of the alter-
definition rules out three classes of decisions where natives and feel that different uncertain events matter
individuals make decisions that affect groups. One to the choice.
class is negotiations, because the individual negotia- 2. Even when group members agree on the relevant
tors are trying to best satisfy their own objectives consequences and events, they may assign different
rather than the group’s objectives, and each individ- probabilities to those events and different values to
ual has veto power. A second class of decisions ruled those consequences.
out is voting situations where all votes are tabulated 3. Each group member’s perspectives of the deci-
to select an alternative according to a prespecified sion and his or her evaluations of the alternatives
1
Keeney: Foundations for Group Decision Analysis
2 Decision Analysis, Articles in Advance, pp. 1–18, © 2013 INFORMS
must both be incorporated in any group decision Table 1 Decision Analysis Assumptions for Individual Decisions
analysis. This implies that the interpersonal compar- (from Pratt et al. 1964)
ison of preferences for the alternatives among mem- Principles of consistent behavior
bers of the group and the relative importance or IT: Transitivity. As regards any set of lotteries among which the decision
power of each individual in the group be recognized. maker has evaluated his or her feelings of preference or indifference,
these relations should be transitive.
Our group decision analysis model is structured from
IS: Substitutability. If some of the prizes in a lottery are replaced by other
the beliefs and preferences of each of the individ- prizes such that the decision maker is indifferent between each new prize
ual members of the decision-making group. A set of and the corresponding original prize, then the decision maker should be
group decision analysis assumptions, analogous to indifferent between the original and the modified lotteries.
those for individual decisions, are developed for this Principles for scaling preferences for consequences and judgments
concerning events
particular group model. Given the group model and
IP: Preferences. The decision maker can scale his or her preference for
these assumptions, a group can analyze any decision any consequence c by specifying a number 4c5 such that he or she
consistent with the principles of decision analysis. would be indifferent between (1) c for certain and (2) a lottery giving a
probability 4c5 chance at c ∗ and a complementary chance at c o .
IJ: Judgments. The decision maker can scale his judgment concerning
2. Background Literature any possible event Ej by specifying a number p4Ej 5 such that he or she
Decision analysis was initially developed for indi- would be indifferent between (1) a lottery with consequence c ∗ if Ej
occurs, c o if it does not, and (2) and a lottery giving a probability p4Ej 5
vidual decisions. Its logical foundations involve
chance at c ∗ and a complementary chance at c o .
constructing a frame to model that decision for sub-
sequent analysis based on expected utility theory,
which integrates the related concepts of probability consequences cnj , n = 11 0 0 0 1 N and j = 11 0 0 0 1 J , that
will result if alternative An is chosen and event Ej then
and utility. The development of each of these con-
occurs.
cepts has occurred over time, and each concept has
multiple contributors. Ramsey (1926) provided the Each consequence includes all things that matter to
initial theory for decision making based upon these the decision maker from the choice of an alternative.
two concepts. Von Neumann and Morgenstern (1947) The assumptions that Pratt et al. (1964) devel-
developed the formal theory of expected utility, which oped to analyze an individual’s decision are given in
was based on so-called objective probabilities. Savage Table 1. The principles of consistent behavior provide
(1954) extended expected utility theory to incorporate the logical foundation for decision analysis. For inter-
probabilities based on a willingness to act, which are preting the substitutability principle, it is important
often referred to as judgmental or personal probabili- to understand that Pratt et al. (1964, p. 356) defined
ties. A concise history of these developments is found the word prize to mean “either a consequence or the
in the work of Raiffa (1968). right to participate in another lottery whose payout
The foundational basis for an operational theory will be a consequence.” Given these two principles
of decision analysis for an individual decision maker and assumption IP, Pratt et al. (1964) prove that the
was fully developed by Pratt et al. (1964). Because decision maker’s preferences over consequences c can
my motivation for this paper is to develop a general be represented by a utility function u that is a positive
decision analysis model to logically analyze group linear transformation of and scaled by
decisions, my analysis uses the individual decision
u4c 5 = 0 and u4c ∗ 5 = 11 (1)
analysis framework of Pratt et al. (1964) as a basis.
Pratt et al. (1964) frames an individual’s decision for where c and c ∗ provide lower and upper bounds on
analysis in the following summary: the decision maker’s preferences for all consequences.
Frame for an Individual’s Decision Problem. An individual From assumption IJ, the p4Ej 5 terms are the proba-
decision maker must choose among a set of alterna- bilities that represent the decision maker’s judgments
tives An , n = 11 0 0 0 1 N , and wants to choose the best about the likelihoods of the various events.
of this set. The decision maker has specified a set of
mutually exclusive and collectively exhaustive events Individual Decision Analysis Theorem. Given
Ej , j = 11 0 0 0 1 J , one of which will occur, and a set of the assumptions for an individual’s decision analysis in
Decision Analysis, Articles in Advance, pp. 1–18, © 2013 INFORMS 3
Table 1, it follows that the decision maker should evaluate and individual’s utilities, respectively. Their interest
alternative An using its expected utility is whether there is any set of such group probabil-
ities and group utilities that would yield a group
U 4An 5 = èj p4Ej 5u4cnj 51 n = 11 0 0 0 1 N 1 (2)
evaluation of alternatives that is consistent with indi-
where alternatives with higher expected utilities are pre- vidual’s common preferences over those alternatives.
ferred and èj means to sum over all j. Seidenfeld et al. (1989) also invoke the weak Pareto
condition. They then prove that there is, in gen-
The past attempts to extend the decision analysis
eral, no compromise group utility function and group
logic to group decisions implicitly assume that each
probabilities for events that can replicate the com-
of the members of the decision-making group accepts
mon preferences of the individual members of the
a commonly held frame of their decision problem.
group.
Notably, and subtly, it assumes that each member
Mongin (1995) is interested in the implications for
of the group (i) is concerned with the same conse-
group decisions of a group of individuals who each,
quences of the alternatives and (ii) considers the same
and collectively as a group, accept the decision anal-
events to be those relevant to the decision. Thus,
the standard approach has assumed that the group ysis (i.e., Bayesian) assumptions. In addition to com-
accepts the assumptions in Table 1 and searched for bining individual group members’ probabilities over
a group utility function and a group set of proba- events into group probabilities and group members’
bilities that can be used in a result analogous to (2) utilities over consequences into group utilities over
to calculate a group expected utility for each of the consequences, he combines group members’ expected
alternatives (e.g., Raiffa 1968). The next three para- utilities for alternatives into group expected utilities
graphs, following the literature review in Keeney and for those alternatives. With an additional assumption
Nau (2011), summarize the work most relevant to this of strong Pareto optimality, Mongin (1995) proves that
paper. there are no possible combinations consistent with the
Hylland and Zeckhauser (1979) investigate the assumptions.
potential usefulness of separately combining the Positive results have been developed for special
individual’s probability assessments into a group cases. Notably, Harsanyi (1955) proved that when the
probability assessment and the individual’s utility expected utility assumptions hold for both individu-
functions into a group utility function. They also als and the group, Pareto optimality holds, and all
assume that the resulting model should be consistent members of the decision-making group have common
with weak Pareto optimality, meaning that if all mem- probability distributions to describe possible conse-
bers of a decision-making group prefer one alterna- quences of the alternatives, then the group’s expected
tive over another, the resulting group decision model utility for an alternative must be a weighted aver-
must yield a higher expected group utility for the age of the individual’s expected utilities for that
Pareto dominating alternative. In addition, they also alternative.
assume that aggregations should not be dictatorial, Harsanyi’s (1955) result is extended to situations
so group probabilities and the group utility function where both probabilities of events and the utili-
should not be identical to those of any individual in ties for consequences can differ among members of
the group. They prove that no group decision model the decision-making group used by Mongin (1998),
is consistent with these requirements. Chambers and Hayashi (2006), and Keeney and Nau
Seidenfeld et al. (1989, p. 225) succinctly summa- (2011). Mongin (1998) and Chambers and Hayashi
rized the problem: “An outstanding challenge for (2006) assumed that the group can specify expected
Bayesian decision theory is to extend its norms utilities for (a) the same alternatives as the indi-
of rationality from individuals to groups.” Their vidual’s consider and (b) a Pareto optimality con-
approach is to broaden the group compromises that dition. Keeney and Nau (2011) assumed that the
they consider to include any set of group probabil- group has expected utility preferences over hypo-
ities and group utilities, which need not be sepa- thetical acts described by lotteries resulting in vector
rately constructed from the individual’s probabilities consequences, where the components are the utilities
perceived by each member of the group, and that the The group alternatives in this paper are the real
group’s preferences are consistent with an assumption alternatives available in the group decision, and the
analogous to Fishburn’s (1965) concept of mutual util- probabilities are descriptions of the group’s judg-
ity independence over these consequences. Because ments over the perceived real consequences that will
the result in Keeney and Nau (2011) is the same as occur to the group. To calculate an expected util-
that for the group decision analysis theorem following ity for the group, assumptions analogous to the
in §5, it is useful to summarize the important distinc- decision analysis assumptions that each member
tions between the theories. These distinctions concern uses to individually analyze the group problem are
the problem formulation, the decision frame, and the used. To relate the group preferences to the mem-
assumptions used. ber’s preferences, an indifference assumption is used.
The origin for the distinction between the result in No Pareto assumption is needed, although the result
Keeney and Nau (2011) and the result here is the prob- is clearly consistent with Pareto optimality.
lem formulation. The problem formulation addressed The current paper has two results, stated in the
by Keeney and Nau (2011) is the standard approach corollary in §5, that are not relevant to the group
where each group member is concerned with the decision frame in the Keeney and Nau (2011) paper.
same set of events and perceives the same conse- With the broader formulation, it is shown that a group
quences for each alternative–event pair. The mem- probability distribution over events acceptable to all
bers may differ in their assignment of probabilities for group members is a product of each of the mem-
events and their utilities for the various consequences. ber’s probability distributions over events in their
The problem formulation of this paper is broader. In individual frame of the group’s decision and that
addition to assigning different probabilities to specific a group utility function over consequences must be
events and different utilities to specific consequences, a weighted average of each of the member’s utility
it allows different group members to be concerned functions over the consequences that each member
with different events and to perceive different con- feels are relevant to the group.
sequences following from any alternative–event pair. The main intent of the two papers is also differ-
Hence, the problem formulation in Keeney and Nau ent. The Keeney and Nau (2011) article is a potential
(2011) is a special case of that addressed here. contribution to the Bayesian expected utility literature
Corresponding to the earlier distinction, the group mainly produced by and of interest to economists.
decision frame is broader in the current paper. The current article is a contribution to the decision
It incorporates each of the member’s decision frames analysis literature. The intent is to provide a logical
of their group decision by essentially using a union of and practical foundation for spreading the applica-
those decision frames. In Keeney and Nau (2011), the tion and use of decision analysis to group decisions.
decision frame for the group decision is exactly the To implement a theory and to have it be useful, it
same as that of each member for their joint decision. is important that the assumptions are stated in terms
Concerning the assumptions used to calculate an that pertain to the real decision being faced, the infor-
expected utility for the group, Keeney and Nau (2011) mation needed to implement the result addresses the
first calculated the expected utilities perceived by complexities of the decision, and practical procedures
each of the individual members for the group’s alter- exist to assess the information. I believe the founda-
natives. To combine these, Keeney and Nau (2011) tions for group decision analysis presented here make
used a decision frame of a hypothetical group deci- such a contribution.
sion problem, where alternatives are defined by
objectively provided probabilities over consequences
described in terms of a vector of the expected util- 3. Frame of the Group Decision
ities that each member perceives for this hypotheti- Problem
cal group decision. Then an assumption analogous to Results described in §2 are important and insight-
Fishburn’s (1965), and a weak Pareto type of assump- ful, but they have not included a comprehensive
tion provide the result. breakthrough that provides the foundations to extend
and implement the decision analysis logic to group Figure 1 Basic Individual Decision Problem
decisions. Thus, if decision analysis is going to be (a) Frame of individual (b) Model of individual
extended to group decisions, it cannot follow the decision decision with all elements
approach of framing the group decision as the indi- E1 p (E1)
c11 u(c11)
vidual decision is framed and then trying to construct
A1 A1
both a group utility function over consequences and
group probabilities for events. A different approach E2 p (E2)
c12 u (c12)
must be followed.
This section describes a different and broader
E1 p (E1)
approach to extend decision analysis to group deci- c21 u (c21)
sions. It incorporates the different frames that group A2 A2
members may have for their common decision, uti- E2 p (E2)
lizes the logic embedded in the decision analysis prin- c22 u (c22)
ciples for individual decisions to analyze the group’s

decision, and produces a result that can be applied
to group decisions using decision analysis techniques To clearly describe the two stages of the group deci-
and procedures. To illustrate different frames, con- sion problem, it is first useful to illustrate the steps
sider a relatively simple decision concerning two peo- in an individual decision analysis model. Figure 1(a)
ple planning to have dinner together at a restaurant. illustrates the frame of an individual decision, for
One individual wants to have a substantial discus- the simplest case of two alternatives and two events,
sion at a convenient location (e.g., not far away and in terms of a decision tree structured directly from
with easy parking), and the other individual is inter- the statement of the individual’s decision problem.
ested in great food at an exciting location. Because It has three elements: alternatives, events, and con-
they have different objectives, they will be concerned sequences. Figure 1(b) presents a model of that deci-
about different consequences of their joint decision. sion that incorporates the decision analysis principles
This will also result in different events being rele- for scaling preferences and judgments. These intro-
vant to each individual, as the first individual would duce two new elements into the model: probabilities
be concerned with how quiet the restaurants are and of events and utilities of consequences.
about travel times to the restaurants, whereas the sec- At the beginning of stage 1 for any group deci-
ond individual may be concerned with whether the sion, the individual members of the group may not
restaurants have exotic fusion dishes and a trendy agree on any of the elements of the group decision.
clientele. The members may view their joint decision as hav-
The decision process for a more complex decision ing different alternatives, events, consequences, prob-
may typically involve several group meetings over abilities, and utilities in terms of Figure 1(b). During
time with individual effort between meetings, as each the meetings, discussions, and information gathering
individual essentially must evaluate the desirabilities of stage 1, each member develops his or her own
of the alternatives for the group. Subsequently, the model of the group’s decision that may account for
group must collectively decide which alternative to the ideas of other members. Fortunately it should be
choose. Hence, it is useful to conceptualize this pro- relatively easy for a group to agree on the set of alter-
cess as two distinct sequential stages. In stage 1, each natives, and, as will be shown, that is the only agree-
group member frames the group’s common decision ment necessary to conduct a group decision analysis.
and evaluates alternatives using the individual deci- If individuals initially identify different sets of alter-
sion model illustrated in §2. In stage 2, the group natives for a joint decision, the union of the individ-
uses the member’s evaluations to collectively evaluate ual alternatives in each of the member’s sets is the
alternatives. This could be done using a discussion, appropriate set of alternatives to use for the group
a voting mechanism, or the more systematic group decision. If some members feel that specific alterna-
decision analysis discussed here. tives suggested by others are inadequate, they have
the possibility and responsibility to evaluate those Figure 2 Individual Frames for Two-Member Group Decision
alternatives as poorly as they feel appropriate. (a) Frame of individual l1 (b) Frame of individual l2
However, even after substantial discussions, the
group members may not agree on either events or E1 F1
c11 d11
consequences that are relevant to the group decision. A1 A1
Accounting for this reality is distinct from essentially
E2 F2
all of the previous work on group decision analysis, c12 d12
which either explicitly or implicitly assumes that all
group members recognize the same events and per- E1 F1
c21 d21
ceive the same consequences. Furthermore, even if
A2 A2
group members happen to agree on the events and
consequences, they may not and need not agree on E2
c22
F2
d22
their judgments of the probabilities of those events or
their utilities for the consequences.
By the end of stage 1, each group member has com- be developed in a mind, and only individuals have
pleted his or her analysis of the group’s alternatives, minds. Groups do not have a mind, so all original
and stage 2 begins. Because each individual’s analysis judgments must occur in the mind of an individ-
has taken into account all of the information relevant ual. However, a group of individuals can decide to
to the group decision (i.e., alternatives, events, con- express group judgments, which leads us to a related
sequences, probabilities, and utilities) from that indi- principle.
vidual’s perspective, at this stage no group member Group Judgments Principle. Any group judgments
should alter his or her analysis in any way because of must be constructed by the group based only on the
any other group member’s information or analysis. judgments of the individuals in the group.
To illustrate the structure of a group decision faced The logic for this principle is as follows: If no
in stage 2, consider the simplest case of a group deci- group member recognizes some aspect that might be
sion that involves two individual group members I1 relevant to a group decision, then clearly the group
and I2 , two common alternatives A1 and A2 , two cannot recognize and consider it. If a potentially rele-
events for each group member, and a consequence vant aspect is recognized by one group member, but
for each alternative–event combination. To keep nota- neither that group member nor any other member
tion simple, suppose that the events of relevance to I1 cares about it, the group should not care about it. Suc-
and I2 are different and labeled E and F , respectively, cinctly, this principle implies that the group should
and the corresponding consequences are different and not care about anything that no member in the group
labeled c and d. The two members’ separate frames cares about, and the group should care about any-
of their joint decision are illustrated in Figure 2. thing that at least one group member does care about.
The frame of the group decision should be con- From these principles, the two group members’
structed from the two group members’ frames of their frames of their joint decision in Figure 2 are combined
group decision. The logic for this is based on the fol- in Figure 3, which illustrates two equivalent frames
lowing principles, which are part of the reason for of the two-member group decision faced in stage 2.
and consistent with the two-stage decision process In the group decision, there are two alternatives A1
described previously. and A2 , four possible joint events represented by
Origin of Judgments Principle. All original judgments (E1 F 5, and eight possible group consequences, rep-
must be developed in the minds of individuals. resented as vectors of the form (c1 d5 expressing the
This principle refers to all judgments such as rec- various possible combinations of consequences in the
ognizing or creating alternatives, describing conse- original decision as perceived by the respective mem-
quences, identifying events, specifying probabilities bers. From Figure 3(a), it is easy to note that if the F
for events, and constructing utilities for consequences. events and d consequences, which are relevant only to
The logic for this principle is that judgments can only member I2 , are eliminated, the group frame reduces to
Figure 3 Equivalent Frames of Two-Member Group Decision
(a) Frame with conditional events (b) Model with joint events
Group Group
probabilities utilities
F1
(c11, d11) pG(E1, F1)
E1 uG(c11, d11)
F2 pG(E1, F2)
(c11, d12) uG(c11, d12)
A1 A1
F1 pG(E2, F1)
(c12, d11) uG(c12, d11)
E2
pG(E2, F2)
F2 uG(c12, d12)
(c12, d12)
F1 pG(E1, F1)
(c21, d21) uG(c21, d21)
E1
F2 pG(E1, F2)
(c21, d22) uG(c21, d22)
A2 A2
F1 pG(E2, F1)
(c22, d21) uG(c22, d21)
E2
F2 pG(E2, F2)
(c22, d22) uG(c22, d22)
that of member I1 shown in Figure 2(a). This indicates An , n = 11 0 0 0 1 N , one of which the group will choose.
that the group decision frame includes a complete The group member has specified a set of mutually
characterization of everything I1 cares about, namely, exclusive and collectively exhaustive events Emj , j =
only E events and c consequences. It follows that if 11 0 0 0 1 Jm , one of which will occur, and a set of conse-
I1 placed any direct value on d consequences, this quences cnmj , n = 11 0 0 0 1 N and j = 11 0 0 0 1 Jm , that will
would involve double counting, because everything result if alternative An is chosen and event Emj then
I1 cares about is already accounted for with the c con- occurs.1
sequences. In addition, I1 would not care about the Each of the group member’s decision frames rep-
probabilities for F events, as these can have no impli- resent their personal perspective of the group deci-
cations for the judgments of I1 about the chances of sion at the time a group decision will be made. Their
different c consequences occurring. The analogous sit- individual decision frames allow them to evaluate all
uation applies for I2 . of the group’s alternatives for input to that group
To analyze the two-member group decision in Fig- decision.
ure 3(b), the group needs joint probabilities for all Figure 5 illustrates the group decision frame.
possible combinations of (E1 F 5 events, represented by The events of interest to the group are members
pG 4Er 1 Fs 5, and group utilities for each of the possible of a set E of mutually exclusive and collectively
(c1 d5, represented by uG 4c1 d5. exhaustive events Eg , g = 11 0 0 0 1 G of the form
The decision frame for the M-member group deci- 4E1j 1 0 0 0 1 Emj 1 0 0 0 1 EMj 5, where Emj , j = 11 0 0 0 1 Jm are the
sion is constructed from each member’s individual events specified by member Im . The consequences of
frames for their group decision analogous to the sim-
1
ple two-member example in Figure 3. An individual’s Because the events relevant to member Im are different in general
from those relevant to any other group member, it may be more
decision frame for their group decision is shown in
appropriate to put a subscript m on all subsequent j notations.
Figure 4 and is defined as follows. Because this would result in subscripts to subscripts, and because
Frame for a group member’s decision. Group member an adjacent subscript m always appears with the subscript j, the
Im , m = 11 0 0 0 1 M, must evaluate the set of alternatives second subscript will not be used.
Figure 4 Decision Frame of Individual Group Member Im for the This group frame includes every group member’s
Group Decision frame for the group decision. To see this, recall that
A1 Em1 group member Im does not care about any events
c1m1 other than Emj or about any aspects of the conse-
quences other than those of the form cnmj . If the events
and consequences without the subscript m, which are
those not of concern to Im , are eliminated from Fig-
ure 5, the problem frame reduces to that for Im shown
in Figure 4.
An Emj
cnmj
4. The Group Decision Analysis
Assumptions
To analyze the group decision in Figure 5, a group
utility function over the consequences (c1 1 0 0 0 1 cM 5
and a joint probability distribution over the rele-
AN EmJ vant events are needed. The group decision analysis
cNmJ assumptions in Table 2, which are analogous to those
for individual decisions in Table 1, provide the logical
basis to derive a solution to the group decision.
interest to the group are described by a vector c = Note that the assumptions of scaling are existence
4c1 1 0 0 0 1 cm 1 0 0 0 1 cM 5, where cm is a consequence that assumptions rather than constructive assumptions as
member Im , m = 11 0 0 0 1 M perceives for the group. in Table 1 for the individual’s decision analysis. The
Frame of the group decision. A decision-making group group utility function is constructed from the indi-
of M members, M ≥ 2, must choose from a set of alter- vidual group member’s utility functions, and the
natives An , n = 11 0 0 0 1 N , one of which will be cho- group probability distribution is constructed from the
sen. One of the mutually exclusive and collectively individual group member’s probability distributions.
exhaustive events Eg , g = 11 0 0 0 1 G will occur, and a Both of these are done as part of the proof of the
consequence cng , n = 11 0 0 0 1 N , g = 11 0 0 0 1 G will result group decision analysis theorem in §5. The reason-
if alternative An is chosen and event Eg then occurs. ableness of the assumptions prior and the resulting
Figure 5 Group Decision Frame for the Group Decision
A1 E11 Em1 EM1

(c111,…,c1m1,…,c1M1)
An E1j Emj EMj

(cn1j,…,cnmj,…,cnMj)
AN E1J EmJ EMJ (cN1J,…,cNmJ,…,cNMJ)
Group probabilities Group utilities

Table 2 Decision Analysis Assumptions for Group Decisions (b) the decision-making group accepts the group deci-
Principles of consistent behavior
sion analysis assumptions in Table 2 for their group deci-
GT: Group transitivity. As regards any set of lotteries among which the sion in Figure 5, and
decision-making group has evaluated its feelings of preference or (c) the decision-making group accepts the group identi-
indifference, these relations should be transitive. cal indifference assumption,
GS: Group substitutability. If some of the prizes in a lottery are replaced then the group expected utility of any alternative An ,
by other prizes such that the decision-making group is indifferent
between each new prize and the corresponding original prize, then the denoted UG 4An 5, is
decision making group should be indifferent between the original and the
modified lotteries. UG 4An 5 = èm wm Um 4An 5
Principles for scaling preferences for consequences and judgments
= èm wm 4èj pm 4Emj 5um 4cnmj 551 (3)
concerning events
GP: Group preferences. The decision-making group can represent its
preferences over the consequences (c1 1 0 0 0 1 cM 5 in terms of a group
where
utility function.
pm 4Emj 5 is member Im ’s1 m = 11 0 0 0 1 M1
GJ: Group judgments. The decision-making group agrees that there
exists a representation of its judgments about the possible occurrence of probability for event Emj 1
any combination of the events (E1 1 0 0 0 1 EM 5 in terms of a joint probability
distribution function. um 4cnmj 5 is member Im ’s1 m = 11 0 0 0 1 M1
utility for consequence cnmj 1
utility function and group probability distributions
are discussed in §6 after presenting the main result. Um 4An 5 is member Im ’s1 m = 11 0 0 0 1 M1
To relate the individual’s preferences to the group’s expected utility for alternative An 1
preferences, we need one additional identity assump-
tion defined as follows. and the wm , m = 11 0 0 0 1 M are scaling factors that sum to
Group Identical Indifference. If some of the prizes one, where 0 ≤ wm ≤ 1.
in a lottery are replaced by other prizes such that
The proof involves several steps. From assump-
the probabilities of all consequences relevant to each
tion (a), each Im can do his or her own analysis
group member in the original and modified lotter-
of the group’s decision, which provides the individ-
ies are identical, then the decision-making group
ual’s probabilities and utilities for the group’s deci-
should be indifferent between the original and modi-
sion. Given assumption (b), the group can analyze its
fied lotteries.
decision, depending on the group’s probability distri-
Note that the group identical indifference (GII)
bution over all possible combinations of events and
assumption is less restrictive than a Pareto assump-
the group’s utility function over consequences per-
tion that would require that the group must be indif-
ceived by each of the group’s members. Then, using
ferent between lotteries when each of the individual
assumption (c), it is shown that the group prefer-
group members was indifferent.
ences for alternatives depend only on the marginal
probability distributions over the set of events, from
5. Group Decision Analysis Theorem which it follows that the group utility function uG
The main result of this paper is the following theorem. must be of the additive form uG = èm wm um , where
Its prescriptive usefulness and procedures to use it wm are nonnegative weighting factors. Hence, the
are discussed in the following sections. expected utility of each real alternative can then
Group Decision Analysis Theorem. Given a be calculated as a weighted sum of the expected
decision-making group of M ≥ 2 members facing a deci- utilities of that alternative from each individual’s
sion with alternatives An , n = 11 0 0 0 1 N , and assum- perspective.
ing that Proof. Given assumption (a), from (1), each group
(a) each member accepts the decision analysis assump- member Im has a utility function that can be scaled by
tions for individual decision-making in Table 1 for his or
her analysis of the group decision in Figure 4, um 4cm 5 = 0 and um 4cm∗ 5 = 11 m = 11 0 0 0 1 M1 (4)
where any two of the cm consequences need not be, sense,” meaning that the group is indifferent between
but may be, the same, and the same circumstance lotteries over multiple attributes (i.e., referring to the
holds for the cm∗ . Also, each group member has a different individual’s perceived consequences as dif-
set of probabilities pm 4Emj 5 for all possible events that ferent attributes in this case) when the marginal prob-
he or she considers can influence the eventual cm . ability distributions over each attribute are identical.
Hence, each group member should evaluate alterna- From Fishburn’s (1965) Theorem 2, it follows that the
tives using his or her own expected utility group utility function uG must be additive, so
Um 4An 5 = èj pm 4Emj 5um 4cnmj 51 uG 4c1 1 0 0 0 1 cM 5 = èm uG 4cm 5 = èm wm uGm 4cm 51 (8)

n = 11 0 0 0 1 N 1 m = 11 0 0 0 1 M0 (5)
where we can scale uG by
A similar result follows from assumption (b). The
uG 4c1 1 0 0 0 1 cM

5=0 and uG 4c1∗ 1 0 0 0 1 cM
∗
5 = 11 (9)
group should evaluate alternatives using
uG 4cm 5 is defined as uG 4c1 1 0 0 0 1 cm−1

1 cm 1 cm+1
1 0 0 0 1 cM ),
UG 4An 5 = èE pG 4E1j 1 0 0 0 1 EMj 5uG 4cn1E 1 0 0 0 1 cnME 51 (6)
m = 11 0 0 0 1 M, uGm 4cm 5 is defined by
where E = 4E1j 1 0 0 0 1 Emj 1 0 0 0 1 EMj 5 is a member of the
uG 4cm 5 = wm uGm 4cm 51 m = 11 0 0 0 1 M1 (10)
set E of all possible combinations of events; cnmE is
the consequence relevant to Im given alternative An is uGm is scaled by
chosen and E occurs; pG is the group probability of
any group consequence (cn1E 1 0 0 0 1 cnME 5 given alterna- uGm 4cm 5 = 0 and uGm 4cm∗ 5 = 11 m = 110001M1 (11)
tive An is chosen and E occurs; and uG is the group
utility for consequence (cn1E 1 0 0 0 1 cnME 5. and the wm are nonnegative scaling factors where
Because of how pG 4E1j 1 0 0 0 1 EMj 5 in (6) is con- 0 ≤ wm ≤ 1, m = 11 0 0 0 1 M. Evaluating (c1∗ 1 0 0 0 1 cM
∗
5 with
structed (see Figure 3), its marginal probability dis- (8) and using (9) and (11) yields
tribution over Emj must be pm 4Emj 5, m = 11 0 0 0 1 M,
which provides the marginal probabilities over the w1 + w2 + · · · + wM = 10 (12)
(cn1E 1 0 0 0 1 cnME 5 for alternatives An , n = 11 0 0 0 1 N . Now
Note from (4) that u1 is scaled from 0 to 1, so
define qG 4E1j 1 0 0 0 1 EMj 5 = çm pm 4Emj 5, so qG is a joint
u1 4c1 5 can be used as a probability. Now, suppose
probability distribution assuming probabilistic inde-
the group faces a choice between (a) an alternative
pendence of all Emj , m = 11 0 0 0 1 M.
described by a group probability distribution yield-
Consider two alternatives Ap and Aq described by
ing (c1∗ 1 c2 1 0 0 0 1 cM

) with a probability of u1 4c1 5 or
different joint probability distribution functions pG
(c1 1 c2 1 0 0 0 1 cM ) with the probability of 61 − u1 4c1 5] and
and qG . As these have the same marginal probabil-
(b) a sure consequence of (c1 , c2 1 0 0 0 1 cM

). The expected
ities over cm , m = 11 0 0 0 1 M, from the group identi-
cal indifference assumption, the group members each utility to I1 of both the lottery and the sure conse-
consider alternatives Ap and Aq to be identical, so quence c1 is u1 4c1 5, so I1 is indifferent between these
the group must be indifferent between alternatives Ap choices. Obviously, each other group member Im ,
and Aq . This has two important implications. First, a m = 21 0 0 0 1 M, is also indifferent. Hence, by group sub-
group probability distribution over events E that all stitutability, the group must be indifferent. Equating
members can agree on is the group utilities yields
pG 4E1j 1 0 0 0 1 EMj 5 = qG 4E1j 1 0 0 0 1 EMj 5 = çm pm 4Emj 50 (7) u1 4c1 5uG 4c1∗ 1c2 10001cM

5+61−u1 4c1 57uG 4c1 1c2 10001cM

5
= uG 4c1 1c2 10001cM

50 (13)
Second, the group’s preferences for alternatives must
only depend on the marginal probability distribu- Substituting (9) and (10) into (13) yields
tions over those consequences. This is Fishburn’s
(1965, p. 38) condition of “independence in the utility u1 4c1 5w1 uG1 4c1∗ 5 = w1 uG1 4c1 50 (14)
Now, substituting (11) into (14) yields concern to Im , m = 11 0 0 0 1 M, and the resulting analy-
sis would be identical for any probability distribution
u1 4c1 5 = uG1 4c1 50 (15) acceptable to the group.
It is useful to recognize that one could replace the
An analogous logic to (15) yields GII assumption in the statement of the Group Deci-
sion Analysis Theorem with a weak Pareto indiffer-
um 4cm 5 = uGm 4cm 51 m = 21 0 0 0 1 M0 (16)
ence (WPI) assumption and the result (3) would still
Substituting (16) into (8) yields hold. The logic is as follows: A weak Pareto indiffer-
ence assumption would state that if all of the group
uG 4c1 1 0 0 0 1 cM 5 = èm wm um 4cm 51 (17) members were indifferent between two alternatives,
then the group should be indifferent between those
which is the form of the group’s utility function over alternatives. If two alternatives had identical prob-
consequences specified by combining the member’s abilities over the consequences of relevance to each
utility functions over the group consequences. group member, then all of the group members would
Substituting (7) and (17) in (6), the expected util- be indifferent between the alternatives. Hence, WPI
ity to the group of an alternative An can be calcu- would imply that the group would be indifferent
lated from between those alternatives, so the GII assumption
would hold, from which the group decision analysis
UG 4An 5 = 6çm pm 4Emj 576èm wm um 4cnmj 57 theorem follows.
= èm wm Um 4An 51 n = 11 0 0 0 1 N 0 (18)
6. Comments on the Result
Substituting into (18) from the individual decision The purpose of this paper is to provide a logically
analysis result in (5) yields sound and practical result that can be used to guide
prescriptive decision making for groups. To exam-
UG 4An 5 = èm wm Um 4An 5 = èm wm 6èj pm 4Emj 5um 4cmnj 571 ine logical soundness, it is necessary to appraise the
n = 11 0 0 0 1 N 1 (19) problem formulation, assumptions, and result of the
group decision analysis theorem. To examine practi-
which completes the proof and indicates that cality, it is necessary to appraise how the theory can
only the marginal probability distributions over cm , be implemented. Logical soundness is addressed in
m = 11 0 0 0 1 M, are required to compute the group this section and implementation is discussed in §8.
expected utility.
6.1. Comments on the Problem Formulation
Corollary 1. A group probability distribution (6) that The problem formulation is based on the two-stage
all group members can agree upon is pG 4E1j 1 0 0 0 1 EMj 5 = decision process described in §3. In the first stage,
çm pm 4Emj 5 as specified in (7). A group utility function each individual group member analyzes the group’s
consistent with the group value judgments must be of decision using the decision analysis framework for
individual decision making. This framework has been
the additive form uG 4c1 1 0 0 0 1 cM 5 = èm wm um 4cm 5 indicated
accepted for more than 40 years as the basis for pre-
in (17).
scriptive decision making under uncertainty for indi-
As demonstrated in the proof of the theorem, the viduals (Savage 1954, Pratt et al. 1964, Howard 1966,
group probability distribution pG incorporates prob- Raiffa 1968).
abilistic independence among the probabilistic judg- The consequences cm of concern to group mem-
ments of the members Im for events Emj , m = 11 0 0 0 1 M. ber Im in his or her own decision frame of the group
It could be that there are probabilistic dependen- decision are meant to include all implications that
cies among these judgments, but no member of the will matter to that group member from the choice of
decision-making group cares about this, because the an alternative. The consequences clearly characterize
chosen pG includes all the probabilistic judgments of what utilities are needed from the individual. They
also indicate the events that matter to that individual, members may strategically misrepresent information
namely, those events that would affect the conse- about their judgments or preferences. Obviously, the
quences that might occur. And these events character- quality and content of the insights potentially avail-
ize what probabilities are needed from the individual. able from any analysis using the group decision analy-
Each individual is the expert on recognizing the con- sis model depends on whether members strategically
sequences of importance in the group decision from misrepresent information. If they do not, the poten-
his or her own perspective and for providing the cor- tial insights guide the choice for the group based on
responding individual probabilities and utilities. balancing perspectives of the members. If a member
During the first stage of any group decision, sig- does misrepresent information, a potential insight is
nificant interaction would likely occur among group recognition that this is the case. Often, members of
members. As a result, each group member will be able the group would know each other, so they can rec-
to account for other members’ perspectives on conse- ognize when another member’s judgments are out
quences, events, probabilities, and utilities. At the end of line with available information or preferences are
of this stage, each group member’s decision frame not consistent with previously stated views. Based
will have a common set of alternatives, but may be on the circumstances, members can appraise whether
different in terms of all other elements (e.g., events, this is because of strategic misrepresentation or a mis-
consequences, probabilities of events, and utilities for understanding of some of the complexities of the
consequences). At this point, each group member, and decision. In applications, group choices about shar-
the group as a whole, are agreeing that the speci- ing the member’s judgments and preferences should
fied consequences, events, probabilities, and utilities be made by considering the transparency of the
of group member Im , m = 11 0 0 0 1 M, are fixed and model, the influence of one member on others, and
appropriately represent his or her judgments and strategic misrepresentation. As with any application
preferences. of a model, the group must subsequently decide
In the second stage, the decision frame for the whether and how the results of the model should
group decision is constructed from all of the group be used in providing insight to help make the group
members’ frames, and when viewed in terms of only decision.
what is relevant to an individual group member, The relative desirability of alternatives from each
reduces to that individual’s frame of the decision. This decision maker’s perspective is described by his
group decision frame, expanded compared to those or her expected utilities of those alternatives. The
used in previous research, accounts for the reality that interpersonal comparison of utilities and the relative
individual group members may perceive that differ- importance of the different members of the decision-
ent consequences and different events are relevant to making group are both incorporated in the scaling
the group’s decision in addition to possible different factors denoted by wm in (3). The decision-making
utilities for the consequences and possible different group must collectively decide on the values of those
probabilities for the events. scaling factors as described in §8.
In the special case where all group members are
concerned with exactly the same consequences and 6.2. Comments on the Group Decision
events, the subscript m for member on consequences Assumptions
cnmj and events Emj could be dropped from the prob- Because each group member has transitive prefer-
lem formulation. If this were done, the main result (3) ences, if the group has expressed its preferences for
reduces to certain lotteries, then it seems reasonable to assume
that the group should wish to make decisions con-
UG 4An 5 = èm wm Um 4An 5 = èm 4wm èj pm 4Ej 5um 4cnj 551 (20)
sistent with the group transitivity assumption GT.
and holds when members have the same or different Note that this assumption does not address situations
probabilities and utilities. where the group has not expressed preferences for
In any group decision, whether or not a model is a lottery, nor does it restrict in any way how group
used to provide insight for making the decision, group preferences are constructed.
The reasonableness of the group substitutability The logic for the group identical indifference GII
assumption GS follows the same logic. Because each assumption is as follows. The original and modi-
group member accepts substitutability, if a group fied lotteries referred to in the assumption deter-
replaces prizes in a lottery with new prizes deemed to mine possible consequences of two alternatives. From
be equal in value to the original prizes, this should not any group member’s perspective, the probabilities of
change the group preferences from that for the origi- the possible consequences of the two alternatives are
nal lottery. This assumption does not specify how the identical. Thus, consistent with the group judgment
group determines that prizes are indifferent, but con- principle, the group as a whole should consider them
cerns only group preferences when substitute prizes to be identical. The group should evaluate two iden-
are indifferent to original ones. tical alternatives, which should be perceived as the
The group judgments assumption GJ is that there same alternative, as equivalent, so they should be
exists a group judgment about the possible occur- indifferent between the two alternatives.
rences of all event combinations (E1 1 0 0 0 1 EM 5 in terms
of a joint probability distribution. The reasonableness 6.3. Comments on the Group Decision Analysis
of this assumption is supported by the following. Result
First, as demonstrated in the proof of the group deci- To contrast use of the group decision analysis theorem
sion analysis theorem, only the marginal probability with previous work, four cases will briefly be consid-
distributions over each of the Em , m = 11 0 0 0 1 M are ered. Case 1, the most general case, is when the only
needed in the proof. Second, from the structure of
thing in common in the members’ frames of the group
the group decision problem, each group member Im
decision is the set of alternatives. Members have dif-
only cares about Em and does not at all care about
ferent consequences, events, probabilities, and utili-
Er 1 r 6= m. Thus, any joint probability distribution over
ties. In this case, there is nothing but the expected
(E1 1 0 0 0 1 EM 5 that has the marginal probability distri-
utilities of the alternatives that could be combined to
butions provided by Im for each Em , m = 11 0 0 0 1 M
evaluate the alternatives from the group’s perspec-
should be acceptable to all group members. Hence,
tive, which is the result of the group decision analysis
the group probability distribution in (7) constructed
theorem.
from the group member’s probability distributions is
Case 2 is a special case of case 1, where the mem-
an appropriate group probability distribution for the
bers agree on the consequences and events, but may
decision.
The group preferences assumption GP is a group disagree on the probabilities for these events and the
commitment to fulfill its collective responsibility to utilities of the consequences. This frame is the one
make a decision in a reasonable manner and not that has been implicitly assumed in most of the previ-
inadvertently select a significantly inferior alternative. ous investigations of foundations for group decisions.
Because each member accepts the decision analysis Consistent with the group decision analysis theorem,
logic for their individual evaluation of alternatives, no agreement on probabilities of events or on utilities
whereas preserving the possible distinctions in events, of consequences is required. The member’s probabil-
probabilities, consequences, and utilities, it is reason- ities and utilities are used in the member’s decision
able to assume a group utility function exists as long analyses, and their expected utilities of the alterna-
as it is of a form that does not require agreements tives are weighted in (3) to provide a group expected
on probabilities and utilities that do not reflect real- utility.
ity. Also, because only group member Im is concerned Case 3 is a special situation of case 2, where the
about consequences cm , it makes sense to use the members also agree on the probabilities of events,
utility function of Im over cm as the group’s utility but may disagree on the utilities of the consequences.
function over cm . Consistent with assumption GP, the In this case, the result is the same as that found by
group utility function that represents the group’s pref- Harsanyi (1955) using a Pareto optimality assumption
erences must be of the additive form (17) with scaling in addition to individual and group decision analysis
factors specified (discussed in §8). assumptions.
Case 4 is the most restrictive situation of a group evaluations of alternatives, and could choose to fol-
decision, where the group members agree on every- low a single individual’s analysis if they collectively
thing: consequences, events, probabilities, and util- decided to do so. Because of this, there seemed to
ities. In this case, the group analyzes the decision be no reason to exclude the dictatorial possibility by
exactly as each of the group members analyze the introducing another assumption.
decision, and the expected utilities of alternatives for
the group are the same as those calculated in each
member’s analysis. 7. Relationship to Arrow’s
Although not assumed, the result of the group deci- Impossibility Theorem
sion analysis theorem is consistent with weak Pareto Arrow (1951, 1963) investigated a specific group deci-
optimality, which is a basic assumption in much of sion formulation to combine the preferences of indi-
economic literature on the group decision making, vidual members of a group for the alternatives to
including the works of Harsanyi (1955), Hylland and obtain a group preference for those alternatives.
Zeckhauser (1979), Seidenfeld et al. (1989), and Mon- In general, he wanted to obtain group preferences P
gin (1995). Clearly, when each group member prefers for all An , n = 11 0 0 0 1 N given the individual member’s
an alternative Ai to an alternative Ak , then each of preferences Pm , m = 11 0 0 0 1 M, so
their expected utilities Um 4Ai 5 must be greater than
Um 4Ak 5, so by (3), the group expected utility UG 4Ai 5 PG 4An 5 = f 4P1 4An 51 0 0 0 1 PM 4An 551 (21)
must be greater than UG 4Ak 5, therefore the group
must prefer Ai to Ak . where f is a function. Arrow’s interest was where
There are no restrictions in this group decision the preferences P were rankings of the alternatives.
framework about whether group members can have He postulated five assumptions and proved that they
common or overlapping events or consequences. For were inconsistent. Hence, an impossibility theorem
a group consequence that is described by (c1 1 0 0 0 1 resulted, which has been very influential and contin-
cm 1 0 0 0 1 cM 5, there may be common aspects in each ues to generate great interest.
cm . On a corporate acquisition decision, each group Suppose one maintains the general formulation
member may include the profit due to any acquisi- (21), but changes the preferences of the individu-
tion as part of the consequences he or she consid- als and the group to be ratings instead of rankings
ers relevant to the group decision. Hence, in each of the alternatives. Then, if one adopts assumptions
cm , m = 11 0 0 0 1 M, that profit is included. Overall, the analogous to Arrow’s using ratings, specifically the
acquisition profit is included M times in the group’s expected utility of alternatives, it has been proven that
consequence. However, because cm is a complete sum- the group decision analysis result (3) follows (Keeney
mary of what matters to Im , um only considers cm , so 1976). Thus, if one assumes that each member of the
there is no double counting of preferences. The point group can specify the expected utilities of the group’s
is particularly clear in the special situation where each alternatives, the result of the group decision anal-
group member is only concerned with profit and has ysis theorem can also be derived from a different
a utility function over profit. Because the sum of the set of logical, but not decision analytic, assumptions.
group member’s weights is one, the group’s weight This is understandable if one recognizes that the for-
on profit is one, so there is no double counting. mulation (21) with expected utilities of alternatives
The group decision analysis theorem allows for a calculated by the individual group members explic-
dictatorial group utility function, meaning one where itly assumes that group’s preferences for alternatives
one scaling factor wm = 1, and all of the others are depend only on the overall ratings (i.e., expected util-
zero. Because the group must collectively specify all ities) of the group members, and implicitly assumes
of the scaling factors, in practice, we would usu- that the group’s preferences for alternatives depend
ally expect each of the wm to be positive. Any group only on the individual member’s marginal probability
will essentially decide how to utilize the member’s distributions over the consequences of alternatives.
8. Comments on Implementation accords one an equal vote. In other situations, the

To implement the group decision analysis result (3) importance of different members may differ depend-
requires information from each of the group mem- ing on characteristics such as ownership, position,
bers and from the group as a whole. Each member or seniority. For example, if three individuals jointly
must verify that the decision analysis assumptions owned an investment with 40%, 35%, and 25% shares,
for an individual decision in Table 1 are appropri- the relative weights for joint decisions may be the
ate for his or her own analysis of the group deci- same as these percentages.
sion and then conduct that analysis. Because these The interpersonal comparison of utilities is a more
are each individual decision analyses, the procedures complex issue that has long been studied and dis-
to do this are well developed and understood (e.g., cussed (e.g., Harsanyi 1955, Luce and Raiffa 1957, Sen
Raiffa 1968, Kirkwood 1997, Clemen and Reilly 2001). 1970, Dyer and Sarin 1979). The essence of the inter-
The group decision model does not require or allow personal comparison of utilities issue is to compare
individuals to assign probabilities to events or utili- the utility to individual I1 of going from a utility of
ties to consequences that are not directly relevant to u1 = 0 to u1 = 1, where u1 = 0 corresponds to the worst
that individual. consequence for the group from the viewpoint of I1
The group as a whole must verify the appropriate- and u1 = 1 corresponds to the best such consequence,
ness of the group decision analysis assumptions in to the utility to individual I2 of going from u2 = 0 to
Table 2 and the group identical indifference assump- u2 = 1. The group may specify equivalent changes in
tion. Verification could occur in a discussion that clar- the utility for I1 and I2 to make this judgment. For
ifies the meaning of each assumption and decides on instance, the group may feel that a change from u1 = 0
whether the group should make their group decision to u1 = 005 is as significant to I1 as a change from
consistent with that assumption. u2 = 0 to u2 = 1 is to I2 . It follows that the change from
The group must also collectively construct the scal- u1 = 0 to u1 = 1 is twice as significant to I1 as a change
ing factors wm , m = 11 0 0 0 1 M to weight the individual from u2 = 0 to u2 = 1 is to I2 . This would imply that
member’s expected utilities in the group evaluation. the ratio of w1 :w2 would be 2:1 if the only relevant
Specifying these weights is perhaps the most difficult concern was the interpersonal comparison of utilities.
aspect of implementing the group decision analysis Suppose this utility comparison was for a decision
framework because (1) it requires value judgments where I1 owned 40% of an investment and I2 owned
about which the individual members may not agree, 60%; the importance contributions to weights were set
and (2) because the sum of the weights must equal at 0.4 and 0.6, respectively. The implication would be
one, the larger the weight given to one individual’s that the relative weights should be 2 × 004 for I1 and
evaluations of the alternatives, the smaller the sum of 1 × 006 for I2 , so the scaling factors would be w1 = 4/7
the weights for the other individuals’ evaluations. and w2 = 3/7, respectively.
A brief description of the conceptual foundation for In practice, if there is a reason for having the rel-
specifying weights will provide a basis for discussing ative importance of group members be different, this
the practical aspects of specifying the weights and issue will likely come up in group discussions and
using the group decision analysis model. Each indi- be explicitly considered. Following the origin of judg-
vidual’s weight depends on two separate factors, the ments principle, it takes one member to think of
relative importance of that individual in the group this concern and bring it to the group’s attention for
and the interpersonal comparisons of group mem- consideration. This may result in selecting specific
ber’s utilities. These two factors can be addressed sep- judgments about relative weights as illustrated ear-
arately, and their implications are easily combined, as lier. If no member brings up the issue of the relative
we will see. importance of group members for the decision, then
In many situations, the group would either explic- it will likely be implicitly assumed that all members
itly or implicitly assume that the importance of all should be treated as equal in this respect.
group members is equal; the basis for such a judg- For some decisions, the interpersonal comparison
ment is simply that being a member of the group of utilities factor may also be unconsciously handled
by implicitly assuming that the changes of utilities For any model to be useful, in addition to being
from um = 0 to um = 1 are equally significant for all Im . able to obtain the information necessary to imple-
If this issue is brought up explicitly, it may be possible ment the model, it must be relevant to some impor-
to specify weights using agreed-upon value trade-offs tant problems. Bordley (2009) discusses classes of
as discussed in the two-person example earlier and important business decisions where the members of
the fact that all weights must sum to one. Baucells a decision-making group would reasonably identify
and Sarin (2003) investigated minimum agreements different events as relevant to the decision. The same
among group members that are sufficient to specify logic for events and consequences is relevant to many
the weights on the member’s utilities when the group governmental and public decisions as indicated by
the following case.
utility function is additive.
As it happens, the group decision analysis
If through open discussion the group does not
model (3) was essentially applied to a very significant
reach an agreement on specifying the set of member’s
decision years ago. Dyer and Miles (1976) worked
weights, one of two procedures may be useful. The
with the Jet Propulsion Laboratory to provide an anal-
first is to have each group member construct a set
ysis of NASA’s Mariner Jupiter/Saturn 1977 project
of scaling factors that he or she feels is appropriate. alternatives. This was a 300 million (in 1977 dollars)
Analysis with each of these sets may provide insights project that included two spacecraft trajectories to be
to reach an agreed-upon decision. launched within a couple of weeks of each other and
The second procedure, which may utilize infor- fly by both Jupiter and Saturn to conduct numerous
mation from the first procedure, is to jointly set scientific experiments.
bounds on the relative importance of group mem- There were 10 groups of scientific studies to be con-
bers and on their interpersonal utility comparisons or ducted on this mission. A different scientific team had
directly on the relative scaling factors, both of which responsibility for each group of studies. A Scientific
lead to bounds on the weights. For example, a group Steering Committee, composed of the leaders of each
may unanimously agree that the significance to mem- of the 10 scientific teams, had responsibility for cre-
ber I1 of going from a utility of u1 = 0 to u1 = 1 is ating and choosing trajectory pairs. Because the sci-
greater than, but not twice as great as, the significance entific experiments concerned distinct fields of study,
to member I2 of going from u2 = 0 to u2 = 1. This such as infrared radiation, imaging science, ultravio-
would indicate that the relative weight of w1 to w2 let spectroscopy, and cosmic ray particles, the teams
must range between one and two. For a two-member naturally had different consequences of concern and
group of equally important individuals, the weight of different events that could affect those consequences.
w1 must be between 1/2 and 2/3, with w2 = 1 − w1 . These characteristics are exactly those assumed in
the group decision analysis frame of this paper, and
The group decision analysis using (3) can be done
the previous standard group decision analysis frame
over the collective ranges of the weights that are
would not be appropriate for a decision with these
acceptable to all group members of the decision-
characteristics.
making group. Depending on how restrictive the
To evaluate alternatives, Dyer and Miles (1976)
weights are and on the specifics of the individual’s
used a few different procedures including what they
expected utilities for the alternatives, the acceptable referred to as an “additive collective choice rule” that
ranges of weights may be sufficient to provide a had intuitive appeal and is essentially the group deci-
unique ordering of the alternatives. In other cases, it sion analysis result (3). They had the scientific teams
is likely that the least desirable alternatives would be each provide the expected utilities for their respective
eliminated from consideration based on the analysis experiments given each trajectory pair. In addition
using the acceptable ranges of weights. If the expected to other sensitivity analyses, the Steering Committee
utilities of the individuals have greater positive corre- selected two sets of scaling factors for the additive
lation, given bounds will tend to provide more defini- collective choice rule and used each to lend insight
tive group evaluations of the alternatives. about the relative desirability of the alternatives. Dyer
and Miles (1976) provided details about how the eval- expected utility for an alternative is the weighted sum
uation process helped in both guiding the Steering of the individual member’s expected utilities for that
Committee to select an alternative trajectory pair and alternative. This result incorporates and maintains the
in honing that pair of trajectories to improve its value. integrity of each member’s decision analysis of what
Their application definitely indicates the potential rel- he or she feels is in the group’s interest and, in addi-
evance of the group decision analysis result to impor- tion, explicitly addresses how the evaluations of the
tant decisions. group members should be combined. The result is a
logically sound operational framework to conduct a
9. Summary decision analysis of any group decision.
There have been several attempts over the past half-
century to extend decision analysis from individual Acknowledgments
The comments of David Bell of Harvard University and
to group decisions. None have resulted in a decision
Robert Nau of Duke University were very helpful and much
analysis framework to analyze the general case of a appreciated.
group decision. These attempts, including those by
Raiffa (1968), Hylland and Zeckhauser (1981), Seiden-
feld et al. (1989), and Mongin (1995), each resulted in References
impossibility theorems. In retrospect, a key contribu- Arrow KJ (1951) Social Choice and Individual Values, (2nd ed. 1963)
(John Wiley & Sons, New York).
tor to the difficulty in finding a positive solution to
Baucells M, Sarin RK (2003) Group decisions with multiple criteria.
the group decision was the implicit assumption in all Management Sci. 49(8):1105–1118.
of these efforts that the group decision had the same Bordley RF (2009) Combining the opinions of experts who partition
frame as an individual decision, which implies that events differently. Decision Anal. 6(1):38–46.
Chambers CP, Hayashi T (2006) Preference aggregation under
the group members are each concerned with the same uncertainty: Savage vs. Pareto. Games Econom. Behav. 54(2):
set of consequences and with the same set of possi- 430–440.
ble events. As a result, the search for a solution pro- Clemen RT, Reilly T (2001) Making Hard Decisions with Decision Tools
ceeded to specify group probabilities for these events (Duxbury, Pacific Grove, CA).
Dyer JS, Miles RF Jr (1976) An actual application of collective
and group utilities for these consequences to produce choice theory to the selection of trajectories for the Mariner
a group decision analysis. As group members could Jupiter/Saturn 1977 project. Oper. Res. 24(2):220–244.
have different judgments about probabilities of events Dyer JS, Sarin RK (1979) Group preference aggregation rules based
on strength of preference. Management Sci. 25(9):822–832.
and different preferences for the consequences, simul-
Fishburn PC (1965) Independence in utility theory with whole
taneously combining both different probabilities and product sets. Oper. Res. 13(1):28–45.
different utilities turned out to be problematic. Harsanyi JC (1955) Cardinal welfare, individualistic ethics, and
The approach taken in this paper addresses the interpersonal comparisons of utility. J. Political Econom.
63(4):309–321.
more realistic general decision problem where group
Howard RA (1966) Decision analysis: Applied decision theory.
members may have different perceptions, and there- Hertz DB, Melese J, eds. Proc. Fourth Internat. Conf. Oper. Res.
fore different decision frames, of their common group (Wiley Interscience, New York), 55–71.
decision. The group decision frame explicitly incorpo- Hylland A, Zeckhauser R (1979) The impossibility of Bayesian
group decision making with separate aggregation of the beliefs
rates each member’s frame, so it is broader than any and values. Econometrica 47(6):1321–1336.
member’s decision frame. This allows each member Keeney RL (1976) A group preference axiomatization with cardinal
to incorporate his or her potentially different conse- utility. Management Sci. 23(2):140–145.
quences and events of concern, as well as different Keeney RL, Nau R (2011) A theorem for Bayesian group decisions.
J. Risk Uncertainty 43(1):1–17.
probabilities of events and utilities of consequences,
Kirkwood CW (1997) Strategic Decision Making: Multiobjective Deci-
into the group decision. Using group decision anal- sion Analysis with Spreadsheets (Duxbury Press, Belmont, CA).
ysis assumptions, analogous to those for an indi- Luce RD, Raiffa H (1957) Games and Decisions (John Wiley & Sons,
vidual decision and relevant only for the specific New York).
Mongin P (1995) Consistent Bayesian aggregation. J. Econom. Theory
group decision constructed from the individual mem-
66(2):313–351.
ber’s frames, provides a decision analysis solution Mongin P (1998) The paradox of the Bayesian experts and state-
for group decisions. This solution is that the group dependent utility theory. J. Math. Econom. 29(3):331–361.
Pratt JW, Raiffa H, Schlaifer R (1964) The foundations of decision research from Massachusetts Institute of Technology. His
under uncertainty: An elementary exposition. Amer. Statist. research interests are in the areas of decision making and
Assoc. J. 59(306):353–375. risk analysis. He has applied such work to important per-
Raiffa H (1968) Decision Analysis (Addison-Wesley, Reading, MA). sonal decisions and as a consultant for private and public
Ramsey FP (1926) Truth and probability. Kyberg HE Jr, Smokler organizations addressing corporate management problems,
HE, eds. Studies in Subjective Probability (Wiley, New York), environmental and risk studies, and decisions involving
Reprinted in 1964.
life-threatening risks. Prior to joining the Duke faculty, Pro-
Savage LJ (1954) The Foundations of Statistics (John Wiley & Sons,
fessor Keeney was a faculty member in Management and
New York).
in Engineering at MIT and at the University of Southern
Seidenfeld T, Kadane JB, Schervish MJ (1989) On the shared
preferences of two Bayesian decision makers. J. Philosophy
California, a research scholar at the International Institute
86(5):225–244. for Applied Systems Analysis in Austria, and the founder of
Sen A (1970) Collective Choice and Social Welfare (Holden-Day, the decision and risk analysis group of a large geotechnical
San Francisco). and environmental consulting firm. Professor Keeney is the
Von Neumann J, Morgenstern O (1947) Theory of Games and author of many books and articles, including Value-Focused
Economic Behavior, 2nd ed. (Princeton University Press, Thinking, Decisions with Multiple Objectives, coauthored with
Princeton, NJ). Howard Raiffa, and Smart Choices, coauthored with John S.
Hammond and Howard Raiffa, which has been translated
Ralph L. Keeney is a research professor emeritus at the into 15 languages. Dr. Keeney was awarded the Ramsey
Fuqua School of Business at Duke University. His edu- Medal for distinguished contributions in decision analysis
cation includes a B.S. in engineering from the Univer- by the Decision Analysis Society and is a member of the
sity of California, Los Angeles, and a Ph.D. in operations U.S. National Academy of Engineering.
View publication stats

Foundations For Group Decision Analysis

Uploaded by

Document Informationclick to expand document information

Copyright:

Available Formats

Foundations For Group Decision Analysis

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Foundations For Group Decision Analysis

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.