Professional Documents
Culture Documents
NOgrH. HOILAND
D. Marc Kilgour
Department of Mathematics
Wilfrid Laurier University
Waterloo, Ontario, Canada N2L 3C5
and
Department of Systems Design Engineering
University of Waterloo
Waterloo, Ontario, Canada N2L 3G1
Liping Fang
Department of Mechanical Engineering
Ryerson Polytechnic University
350 Victoria Street
Toronto, Ontario, Canada M5B 2K3
and
Department of Systems Design Engineering
University of Waterloo
Waterloo, Ontario, Canada N2L 3G1
and
Xiaoyong (John) Peng
Department of Systems Design Engineering
University of Waterloo
Waterloo, Ontario, Canada N2L 3G1
ABSTRACT
The graph model for conflict resolution is put forward as a unique decision
technology for systematically studying strategic conflicts that can arise in environ-
mental management and elsewhere. This flexible conflict resolution methodology has
solid and realistic mathematical foundations which permit it to model strategic
decisions accurately, forecast compromise solutions, and assist in assessing the
political, economic, environmental, and social viability of alternative scenarios to
resolving environmental conflicts. To permit practitioners to conveniently apply the
graph model for conflict resolution to practical real-world conflict problems, a new
decision support system called GMCR II (Graph Model for Conflict Resolution II) is
designed. To explain the basic structure of GMCR II and how it works in practice,
this paper describes its use to model and analyze an international environmental
management dispute that took place in North America. © Elsevier Science Inc.,
1997
INTRODUCTION
The recently developed graph form of the conflict model extends and refines
these methodologies to describe more accurately the behavior of participants
in a strategic conflict [5-7]. The graph form takes states, rather than
individual decisions, as the basic units for describing a conflict. A rich
variety of possible social behavior in a conflict can then be represented by
appropriate solution concepts. This paper focuses upon conflict resolution
methodologies developed by the authors and their colleagues. Overviews of
game-theoretic models and their use in decision making, plus key references
for various decision models, are provided in [6, 8].
In the next section, the theory underlying the graph model for conflict
resolution is outlined. Subsequently, the option form for representing a
conflict is explained using an international resource dispute. This conflict
was occasioned by a proposed open-pit coal mining development in the
province of British Columbia, Canada, along a tributary of the Flathead
River, which flows into the state of Montana. Because effluent from the
mining operation would have adverse effects downstream, especially on
recreational water use, the development faced international opposition from
the federal government of the United States and the state government of
Montana. In accordance with the Boundary Waters Treaty of 1909, the
governments of the United States and Canada requested the International
Joint Commission (IJC) to examine the entire situation.
In the following section, a general design for the new decision support
system GMCR II is presented. Its objective is to allow practitioners to
conveniently and expeditiously apply the graph model approach to an actual
conflict. To demonstrate the efficacy of the graph modeling approach to
conflict resolution in resource development, GMCR II is employed for
modeling and analyzing this international environmental conflict at a crucial
point in its evolution, just before the IJC made its recommendation. In fact,
the conflict study presented in the paper constitutes a revised and expanded
version of earlier analyses carried out in [9, 10]. The conference paper [9]
contains material about GMCR II complementary to what is presented here.
exercise real power in a conflict. The next two subsections outline some of
the key ideas behind modeling and stability analysis, respectively. Following
a description of how to model a strategic conflict in option form, the
computer implementation of these ideas in the decision support system
G M C R II is described; then G M C R II is applied to the environmental
conflict.
Modeling
The graph model for conflict resolution represents a conflict as moving
from state to state (the vertices of a graph) via transitions (the arcs of the
graph) controlled by the DMs. One inherent advantage of the graph model is
that it can incorporate irreversible moves, whereby a DM can unilaterally
move from state k to state q but not from q to k. Examples of irreversible
moves are represented later in the case study. Another key advantage of the
graph model is its ability to easily describe c o m m o n moves, in which more
than one DM can cause the conflict to move from one particular state to
another common state. For instance, the greenhouse effect can be caused by
greenhouse gases emitted from any one of m a n y countries around the world.
A graph model for a conflict consists of a directed graph and a payoff
function for each DM who can affect the dispute. Let N = {1, 2 , . . . , n}
denote the set of DMs and U = {1, 2 , . . . , u} the set of states or possible
scenarios of the conflict. A collection of finite directed graphs D i = (U, A i),
i ~ N, can be used to model the course of the conflict. The vertices of each
graph are the possible states of the conflict and therefore the vertex set, U,
is common to all graphs. If DM i can unilaterally move (in one step) from
state k to state q, there is an arc with orientation from k to q in A 4. In the
next main section, it is shown how the option form can be used in G M C R II
to conveniently generate the states and the graphs in a model.
DM i's graph can be represented by i's teachability matrix, R i , which
displays the unilateral moves available to DM i from each state. For i ~ N,
R , is the u × u matrix defined by
R,(k, k) = 0.
S i(k) of all states to which DM i can move (in one step) from state k.
Therefore,
S i ( k ) = {q ~ U: Ri(k, q) = 1}.
s : ( k ) = {q s,(k): R : ( k , q) = 1}.
Stability Analysis
The stability analysis of a conflict is carried out by determining the
stability of each state for every DM. A state is stable for a DM iff that DM
has no incentive to deviate from it unilaterally, under a particular stability
definition or solution concept. A state is an equilibrium or possible resolu-
tion under a particular solution concept iff all DMs find it to be stable under
that stability definition.
are many different solution concepts. Fang et al. define [6, Chapter 3] and
mathematically compare [6, Chapter 5] the solution concepts listed in Table
1. Furthermore, they demonstrate how the graph model and its associated
solution concepts can be equivalently expressed as extensive form games
[6, Chapter 4].
The solution concepts in Table 1 can be applied to conflicts having two or
more DMs. The first column gives the names of the solution concepts and
their acronyms, while the second one provides original references. The last
three columns furnish qualitative characterizations of the solution concepts
according to the three criteria of foresight, disimprovements, and knowledge
of preferences. Foresight refers to the ability of a DM to think about
possible moves that could take place in the future. A DM who has high or
long foresight can imagine many moves and countermoves into the future
when evaluating where the conflict could end up because of an initial
unilateral move on his or her part. Notice, for example, that in Nash
stability foresight is low, whereas, it is very high for nonmyopic stability.
The criterion of strategic disimprovements in the fourth column means that
a particular DM may temporarily move to a worse state in order to reach a
more preferred state eventually. Disimprovements by opponents indicate
that the other DMs may put themselves in worse positions in order to block
unilateral improvements by the given DM. The last criterion, knowledge of
preferences, refers to the preference information used in a stability analysis.
For example, under R, GMR, or SMR, a DM need not know the preferences
of all other DMs, although their ability to move unilaterally to other states
TABLE 1
SOLUTION CONCEPTS AND HUMAN BEHAVIOR
Knowledge of
Solution Concepts References Foresight Disimprovements Preferences
Nash Stability (R) [11, 12] Low Never Own
General [2] Medium By opponents Own
metarationality
(GMR)
Symmetric [2] Medium By opponents Own
metarationality
(SMR)
Sequential stability [3, 4] Medium Never All
(SEQ)
Limit-move stability [7, 13, 14] Variable Strategic All
(Lh)
Nonmyopic stability [7, 13, 15, 16] High Strategic All
(NM)
Environmental Conflict Management 123
Unilateral Moves
For a conflict involving more than two DMs, one must consider sequen-
tial moves involving more than one DM. Let H ___N be any subset of the
DMs. Then SH(k) denotes the set of all states that can result from any
sequence of unilateral moves by some or all of the DMs in H, starting at
state k. In this sequence, the same DM may move more than once, but not
twice consecutively. If all of the DM's graphs are transitive, there is always
an equivalent single move to any sequence of moves in which one DM moves
twice consecutively. The formal definition of a unilateral move by H c N is
COMMON MOVES. A common moves takes place when either of two (or
more) DMs can independently make a unilateral choice that causes the game
to move from one or more initial states to one specific final state.
R E P R E S E N T A T I O N OF C O N F L I C T S IN O P T I O N F O R M
• e
BRITISH
COLLIMB1,4
,CANADA
U.S.A.
TABLE 2
THE DMS AND OPTIONS OF THE FLATHEAD RIVER
RESOURCE DEVELOPMENT DISPUTE
TABLE 3
STATES OF THE F L A T H E A D R I V E R R E S O U R C E D E V E L O P M E N T D I S P U T E
S~eCreek
1. Continue Y Y N N Y N N Y N N Y N N Y N N Y N N
2. M o d ~ N N Y N N Y N N Y N N Y N N Y N N Y N
British
Columbia
3. Original Y Y Y Y N N N N N N Y Y Y N N N N N N
4. M o d ~ c a t i o n N N N N Y Y Y N N N N N N Y Y Y N N N
Montana
5, Oppose Y N N N N N N N N N Y Y Y Y Y Y Y Y Y
IJC
6. Original N Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y
7. M o d ~ c a t i o n N N N N N N N N N N N N N N N N N N N
8. No N N N N N N N N N N N N N N N N N N N
State Numbers 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
Sage Creek
1. Continue Y N N Y N N Y N N Y N N Y N N Y N N
2. Modify N Y N N Y N N Y N N Y N N Y N N Y N
British
Columbia
3. Original Y Y Y N N N N N N Y Y Y N N N N N N
4. Modification N N N Y Y Y N N N N N N Y Y Y N N N
Montana
5. Oppose N N N N N N N N N Y Y Y Y Y Y Y Y Y
IJC
6. Original N N N N N N N N N N N N N N N N N N
7. Modification Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y
8. No N N N N N N N N N N N N N N N N N N
State Numbers 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37
Sage Creek
1. Continue Y N N Y N N Y N N Y N N Y N N Y N N
2. Modify N Y N N Y N N Y N N Y N N Y N N Y N
British
Columbia
3. Original Y Y Y N N N N N N Y Y Y N N N N N N
4. Modification N N N Y Y Y N N N N N N Y Y Y N N N
Montana
5. Oppose N N N N N N N N N Y Y Y Y Y Y Y Y Y
IJC
6. Original N N N N N N N N N N N N N N N N N N
7. Modification N N N N N N N N N N N N N N N N N N
8. No Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y
State Numbers 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55
130 K . W . HIPEL ET AL.
TABLE 4
REACHABLE LISTS [ S i ( ' ) ] AND PAYOFFS [ P~(-)] FOR THE FLATHEAD RIVER
RESOURCE DEVELOPMENT DISPUTE
2 3, 4 55 5, 8 55 11 1 2
3 2, 4 48 6, 9 41 12 20 2
4 2, 3 36 7, 10 45 13 39 2
5 6,7 18 2,8 12 14 3 2
6 5, 7 42 3, 9 4 15 22 2
7 5, 6 30 4, 10 6 16 41 2
8 9, 10 16 2, 5 24 17 5 2
9 8, 10 17 3, 6 14 18 24 2
10 8, 9 24 4, 7 18 19 42 2
11 12, 13 54 14, 17 52 2 2 2
12 11, 13 47 15, 18 39 3 21 2
13 11, 12 35 16, 19 43 4 38 2
14 15, 16 15 11, 17 10 5 4 2
15 14, 16 41 12, 18 2 6 23 2
16 14, 15 29 13, 19 5 7 40 2
17 18, 19 13 11, 14 22 8 6 2
18 17, 19 14 12, 15 13 9 25 2
19 17, 18 23 13, 16 16 10 43 2
20 21, 22 53 23, 26 3 29 7 2
21 20, 22 46 24, 27 11 30 26 2
22 20, 21 34 25, 28 8 31 45 2
23 24, 25 12 20, 26 40 32 10 2
24 23, 25 40 21, 27 54 33 28 2
25 23, 24 28 22, 28 49 34 47 2
26 27, 28 10 20, 23 27 35 12 2
27 26, 28 11 21, 24 37 36 30 2
28 26, 27 22 22, 25 33 37 49 2
29 30, 31 52 32, 35 1 20 9 2
30 29, 31 45 33, 36 9 21 27 2
31 29, 30 33 34, 37 7 22 44 2
32 33, 34 9 29, 35 38 23 11 2
33 32, 34 39 30, 36 51 24 29 2
34 32, 33 27 31, 37 47 25 46 2
35 36, 37 7 29, 32 26 26 13 2
36 35, 37 8 30, 33 35 27 31 2
37 35, 36 21 31, 34 31 28 48 2
Environmental Conflict Management 131
TABLE 4
(CONTINUED)
38 39,40 51 41,44 17 47 14 2
39 38,40 44 42,45 20 48 32 2
40 38,39 32 43,46 23 49 51 2
41 42,43 6 38,44 29 50 16 2
42 41,43 38 39,45 32 51 34 2
43 41,42 26 40,46 36 52 53 2
44 45,46 4 38,41 44 53 18 2
45 44,46 5 39,42 48 54 36 2
46 44,45 20 40,43 53 55 55 2
47 48,49 49 50,53 15 38 15 2
48 47,49 43 51,54 19 39 33 2
49 47,48 31 52,55 21 40 50 2
50 51,52 3 47,53 28 41 17 2
51 50,52 37 48,54 30 42 35 2
52 50,51 25 49,55 34 43 52 2
53 54,55 1 47,50 42 44 19 2
54 53,55 2 48,51 46 45 37 2
55 53,54 19 49,52 50 46 54 2
()
DECISION MAKERS
SC Sage Creek
BC British Columbia
MT _ _ Montana
IJC International
Joint Commission
UC
* ' C STATES 2 - 19
IIC
C ' ~ S 20 - 37 ~.~
UC
...... MT . . . . . . M . c ......
•. " ~ . ....... ~:::'" : ..... i............
. . . . . . . . BC . . . . - . . . . . BC- --"
F r o . 2. The graph model for the Flathead River resource development dispute.
MODEL SUBSYSTEM
Algorithms for
Modeling,
Individual Stabilities,
Solution Concep~, ~ f ] and Equilibria
DATA SUBSYSTEM
• Transform Input Decision Makers
• Retrieve Data Options
• Interpret Output Feasible States
State Transitions
Preferences
Stability Results
GMCR II
DECISION MAKERS
decision moker's
~/~~plan
--stability-type
,,°,. 0,o°.
555 J /
P°7-//
STATES
FIG.4. Organizationof stability results in the data subsystem.
2. For each stability type, the stability-type plane (parallel to the DECI-
SION M A K E R / S T A T E plane) provides a complete analysis of the model
according to that stability type.
3. For each state, the state plane (parallel to the DECISION M A K E R /
STABILITY T Y P E plane) identifies the DMs for whom the particular state
is stable (s), or unstable (u) under each possible stability type.
4. The STABILITY T Y P E / S T A T E plane itself, referred to as the
equilibrium plane, contains the projection of stability results for each DM ( E
if s for all DMs, blank otherwise), indicating all equilibria for each stability
type.
GMCR II is being developed using Borland C ++ within the Microsoft
Windows 3.1 environment. In this way, the system will be widely accessible,
be user-friendly, and possess powerful graphical capabilities. Table 5 fur-
nishes the menu for the User Interface for GMCR II. As can be seen, GMCR
II can employ the option form for defining states, as illustrated below. The
option form is especially useful when the analyst can express the specific
courses of actions as options or general choices available to each DM that
may be selected more or less independently. Because a large number of
states can be automatically calculated by GMCR II using a relatively small
number of options, the option form can be used with very large conflicts.
In some situations, one may prefer to discuss a conflict in terms of overall
states and not worry about the detailed definition of each state. For
example, senior decision makers may be interested in possible final results,
136 K . W . HIPEL ET AL.
TABLE 5
MENUTREEFOR GMCR II
1st Level 2nd Level 3rd Level
File New
Open
Save
Save As ...
Exit
Edit Undo
Cut
Copy
Paste
Model Input Option Form Input DMs/Options
State Elimination
State Combination
Common Moves
Irreversible Moves
Forcing Moves
Graph Form
Preview Conflict Description
Feasible State List
Reachable Lists
Preferences Option Based Preference Tree
Option Weighting
Further Ranking
Graph Based
Stability Output Option Form
Graph Form
Interpretation Status Quo Analysis
Graphical Evolution Path
Further Analysis Sensitivity Analysis
Hypergame Analysis
Coalition Analysis
Help Contents
About...
and in general, how the more preferred results can be achieved. The
graphical input of states mentioned in Table 5 may be very helpful in these
cases, especially for smaller conflicts at early stages of development. A circle
or node can be drawn on a screen to represent a state while possible moves
between states are depicted using arrows. This graphical information can
then be converted by GMCR II into a form which can be processed by the
analysis engine. For output, the user will be able to select different aspects
Environmental Conflict Management 137
Option-based Modeling
GMCR II can employ the option form to input states and state transi-
tions for a model. This information is transformed to become the input for
the state-based analysis engine.
This section presents the design of option-based modeling and shows how
the input information that the analysis engine needs, such as the feasible
states, reachable lists, and preferences, can be generated from the modeling
information. State removal and preferences trees were implemented in
DecisionMaker [37]. These algorithms will be redesigned using the data
structure of GMCR II. Features like state combination, irreversible moves,
common moves, and forcing moves have never previously been implemented
in option form. As much as possible, option form input will be introduced in
the context of the Flathead River dispute, described above.
DMs and Options. Figure 5 shows the prototyping dialog box for DM
and option input, where "British Columbia" in the Decision Makers listbox
is highlighted and, accordingly, the lower input area is ready for British
Columbia's options. The system will read in the information appearing in
this dialog box and store it in a proper format in the data subsystem. If, for
example, the user deletes the highlighted item "British Columbia", the
options belonging to "British Columbia" will also disappear from the option
list as well as from the inner storage, and the label for the rest of the
DMs/Options will be updated accordingly.
Data Structure. The option form can cause problems in memory and
execution time. For example, the DSS I N T E R A C T mentioned above has
difficulty obtaining a list of all feasible scenarios, even in models with only
10 options [35]. Efficient algorithms and data structures based on the
preference tree concept have been devised [41], creating significant savings in
memory and execution time during a stability analysis using the sequential
stability solution concept. In GMCR II, a 32-bit DOUBLEWORD is used to
represent a specific option selection defining a state. This approach is
efficient and can be used in analyses employing any of the solution concepts
given in Table 1. Based on this data structure and the modeling algorithms
introduced below, all feasible states in a model can be generated very
quickly.
138 K.W. HIPEL ET AL.
GMCR - FLATHEAD.GMC
For each digit or bit, a " 0 " or "1" indicates whether the option is
selected by the DM controlling it, or not selected, respectively. Because
there are 32 bits, this format can handle up to 32 options, which is more
than sufficient for use in most real-world applications. If there are n options
in the model, GMCR II automatically works with the resulting 2 n mathe-
matically possible states.
For example, state 1 of the 8-option Flathead River conflict shown in
Tables 2 and 3 is represented in binary form as
32 b~ts
. . . . . . 000 1 01 01 (binary).
DM4 DM3 DM2 DMI
Recall that there are four DMs in this model. The brackets indicate which
options each DM controls.
This data representation is straightforward and analogous to option form
itself. As well, this data structure utilizes the C language's bit-wise operator
features.
Environmental Conflict Management 139
GMCR-FLATHEAD.GMC 1
is not allowed. In other words, "l" must have lowest priority. Here "&"
means "and", and "l" means "or".
140 K. W. HIPEL ET AL.
TABLE 6
ERROR PREVENTING FEATURES WITH STATE ELIMINATION DIALOG BOX
TABLE 7
INFEASIBLESTATESPECIFICATIONSFOR THE FLATHEADRIVERCONFLICT
Input
Information Interpretation
1&2 Sage Creek cannot do both "Continue" and "Modify"
3&4 British Columbia cannot do both "Original" and "Modification"
6&7 IJC cannot do both "Original" and "Modification"
7&8 IJC cannot do both "Modification" and "No"
8&:6 IJC cannot do both "No" and "Original"
information and puts an "(option 1) and (not option 5)" into the "Infeasible
Situation" list, which in turn requires all states with the pattern
"" . . . . . . . . 1 - - - i f ' be eliminated. Here, the dash ..... indicates that the
entry can be either a 0 or 1.
This idea of pattern is actually used to represent a group of states
conveniently and compactly. But a pattern cannot be conveniently recorded
in a simple data structure. However, thanks to the C language's low-level
bit-wise operation, a pair of masks, which are extremely simple to record
and process in the program, can be used to equivalently represent a pattern.
The following two " m a s k s " are used to represent the pattern
......... 1 - - - 0 : m0 . . . . . . . 11111110; m 1 . . . . . . . 00010000. Here
"equivalently" means that if a state k satisfies
then k must belong to the above pattern. (Here "l" and "&" mean bit-wise
OR and bit-wise AND, respectively.) This kind of pattern and the equiva-
lent mask pair will be used throughout the design.
From the state elimination dialog box, a list of patterns can be recorded
and will be used as "filters" for state removal. Infeasible state removal has
been called "one of the most tedious tasks when performing a conflict
analysis" [4, p. 159], and several programs have been developed to deal with
it based on different data structures. In this design, state removal is simple,
efficient, and straightforward.
State Combination and Common Moves. Sometimes in a conflict model,
a group of states with a common pattern is indistinguishable and, thus,
should be treated as a single state. To specify which states are to be
coalesced to form a state combination, users are required to provide the
common characteristics of the states.
142 K . W . HIPEL ET AL.
Common moves happen when either of two (or more) DMs can unilater-
ally force the model to change from one state to another. (Common moves
do exist in the real world. An example in a simplified model of a superpower
nuclear confrontation is given in [6, Chapter 2].) Common moves can be
specified in almost the same way as state combinations. Compared with the
specification of indistinguishable states, the only difference is that common
move specification allows the use of the logic operator OR on cross-DM
options, which is not available for state combinations.
Generation of the Feasible State List. After a user provides the input
information consisting of the DMs, their options, the infeasible states, the
indistinguishable states, and the common moves, GMCR II automatically
generates the feasible state list, which is then available to the user under the
Preview menu. Figure 7 shows how the feasible state list is obtained, where
the filtering is accomplished as described in the section of state removal.
Irreversible Moves. In an irreversible move, a DM can unilaterally cause
a conflict to move from state k to q by changing his or her option selection,
but cannot make the reverse transition from q to k. Accordingly, within the
Irreversible Move dialog box, a user can highlight the option or options
which cause the irreversible move. Figure 8a shows a prototyping design of
this dialog box to specify irreversible moves based on single option status
shift constraints. The information shown in the dialog box is sufficient for
many models. But, in some situations, status shift constraints based on
Interface
generate All mathemaucaUy
I DM/OptionInput
Dialog Box possiblestates I
State Elimination interpret J Infeasible State]
Filter all the states
User
DialogBox "] PaRems ] matching the pattern
.........................
I State Combination
Dialo~Box
Filterall the states matching
I ~. - ~ Indistinguishable] the pattern from the rest
State Patterns I ..........................
Common Moves
Dialo~Box Addeach pattern
into the list as
representative of its group
List of
feasible states
(a)
¢o)
Fic. 8. Irreversible Move dialog box: (a) single option; (b) multiple options.
144 K . W . HIPEL ET AL.
where P ' and P" are simple preference statements. Obviously, the input
information for a set of preference statements can be elicited in a similar
way, as discussed in the section on state removal. Each preference statement
expresses a logical relationship among options t h a t can be used to bifurcate
a set of states. Therefore, a set of lexicographic preference statements can
produce a ranking of the states. An algorithm modified from the QuickSort
algorithm [44] is designed to carry out preference tree ranking based on
G M C R II's d a t a structure. To obtain the ranking based on n preference
statements, states t h a t match the first statement are assigned a weight of
2 ~, the second 2 ~- 1 and so on; states are then ranked according to their
total weights, producing ordinal payoffs which are consistent with corre-
sponding preference tree rankings.
Another general and flexible method is the option weighting method, in
which weights (positive, negative, or zero) are assigned to each option, and
Environmental Conflict Management 145
l ieN;k,qeUI
Doesk
sCOntrol,are the N match a forcing
tatus of k and q pattern
the irreversible N
moves' startpattern
I q~si~k~ I I q~si~ I
I 1
ANALYSIS OF T H E F L A T H E A D R I V E R R E S O U R C E
DEVELOPMENT DISPUTE
TABLE 8
EQUILIBRIA FOR THE FLATHEAD RIVER RESOURCE DEVELOPMENT DISPUTE
k Equilibrium
3 GMR, SMR
4 GMR, SMR
11 R, GMR, SMR, SEQ, L1, L2, L~, NM
12 GMR, SMR
13 GMR, SMR
24 GMR, SMR
25 GMR, SMR
33 R, GMR, SMR, SEQ, L1, L2, L3, NM
34 GMR, SMR
46 R, GMR, SMR, SEQ, L1, L2, L3, NM
55 GMR, SMR
Environmental Conflict Management 147
concepts are taken from Table 1. This table indicates that states 11, 33, and
46 are equilibria for all solution concepts and possess a much greater degree
of stability than any other states. In Tables 8 and 1, the symbol L~ means
that the corresponding state is an equilibrium according to the limited move
stability for a horizon of h steps.
The status quo at the time of analysis is state 1, shown in Tables 2 and 3,
in which Sage Creek is proceeding with the original project, the government
of British Columbia has approved-in-principle Stage II of the project by
supporting the original project, Montana is opposing the project, and the
IJC has not yet made a recommendation. Equilibria 11, 33, and 46 corre-
spond to the IJC's three possible recommendations. If it recommends
stopping the project, the likely resolution is state 46. The actual outcome
was that IJC recommended stopping the project and the immediate result
was state 47. The sequence of state transitions from the status quo state 1
through the transition states 47, 53, and 55 to the equilibrium 46 is shown in
Table 9, where an arrow indicates which strategy selection changes at each
transition.
A sensitivity analysis was conducted by using GMCR II to analyze
alternative models to the one defined in Table 4. These models differ only in
certain aspects of the DMs' preferences. The objective was to assess the
effects of these preference variations on the results of the analysis.
First, British Columbia's preferences were changed to reflect the hypoth-
esis that even if the IJC recommends stopping the project, British Columbia
TABLE 9
STATE TRANSITIONS FROM STATUS QUO TO FINAL OUTCOME
Sage Creek
1. Continue Y Y Y ) N N
2. Modify N N N N N
British Columbia
3. Original Y Y N N N
4. Modification N N N N N
Montana
5. Oppose Y Y Y Y -~ N
IJC
6. Original N N N N N
7. Modification N N N N N
8. No N --* Y Y Y Y
State Numbers 1 47 53 55 46
148 K. W. HIPEL ET AL.
TABLE 10
EQUILIBRIA FOR THE SENSITIVITY ANALYSIS
k Equilibrium
3 GMR, SMR
4 GMR, SMR
11 R, GMR, SMR, SEQ, L1, L2, L3, NM
12 GMR, SMR
13 GRM, SMR
24 GMR, SMR
25 GMR, SMR
33 R, GMR, SMR, SEQ, Lx, L2, L3, NM
34 GMR, SMR
42 GMR, SMR
43 GMR, SMR, SEQ
46 GMR, SEQ, L3, L4, Lb, L6, NM
51 GMR
52 GMR
55 GMR, SMR, SEQ, L2, L3, LI4t, L~l, LIIb,NMH
CONCLUSIONS
GMCR II constitutes the next generation of a DSS for the efficient and
effective management of strategic conflicts. Such conflicts can take place in
land, water, and environmental management, as well as many other areas of
human endeavor. As demonstrated by the real-world resource management
conflict in the previous sections, GMCR II can provide decision makers and
analysts with decision advice, structural insights and answers to what-if
questions. With this enhanced understanding, analysts can better explain
strategic relationships and assist DMs, who may thus gain the opportunity
to direct the evolution of the conflict toward more favorable results. In order
to approach solutions to resource management problems that axe both
economically and environmentally sustainable, it is essential that the view-
points of competing interest groups be taken into account through a decision
technology such as GMCR II.
REFERENCES