You are on page 1of 14

IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 34, NO.

4, JULY 2004 507

Preference Uncertainty in the Graph Model
for Conflict Resolution
Kevin W. Li, Keith W. Hipel, Fellow, IEEE, D. Marc Kilgour, and Liping Fang, Senior Member, IEEE

Abstract—A new preference structure is introduced into the Theory of Games and Economic Behavior [13]. Since its publi-
graph model for conflict resolution. This structure can handle a cation, and especially after the innovative concept of Nash equi-
decision-maker’s (DM) strict preference for one state or scenario librium [14], [15] was proposed, game theory has penetrated
over another, equal preference for states, and uncertain or un-
known preference in the comparison of two states. Built upon this many fields, including economics, psychology, biology, engi-
preference structure, four types of solution definitions modeling neering and politics [2], [6], [16], [17].
human behavior under conflict are extended to accommodate un- The simple idea of modeling a conflict in terms of DMs,
certainty in preferences. Four distinct ways to consider uncertain strategies, states, each DM’s preferences among states, and
preference information are identified, producing sixteen extended strategic interactions among DMs, is the basis of the theoretical
stability definitions. Interrelationships of these definitions within
and across the four definition sets are investigated. Illustrative development of conflict resolution methodologies. Among the
examples of two-DM and multi-DM conflict models are presented aforementioned methodologies to analyze strategic conflicts,
to show how the new solution concepts can be applied in practice. the graph model is probably the simplest and most flexible [2].
Index Terms—Conflict, graph model, preference uncertainty, so- It takes feasible states, or possible scenarios of a conflict, as
lution concepts. basic components and tracks each DM’s possible moves among
them. In the graph model, the stability of each state is assessed
from each DM’s viewpoint by examining potential moves and
I. INTRODUCTION countermoves by other DMs. States that are stable for all DMs
in the conflict constitute equilibria, and correspond to potential
C ONFLICT may arise whenever human beings with
clashing interests and objectives interact with each other
[1]; strategic conflict is commonly observed in virtually all
resolutions of the conflict under study.
Whether a state is stable for a given DM depends on the so-
human activities ranging through bargaining in marketplaces, lution concept selected. Solution concepts characterize DMs’
labor-management negotiations, trade disputes, and even wars. possible behavioral patterns in conflicts. To depict a diversity
The pervasiveness of conflict creates a great need for formal of decision types, a variety of solution concepts has been put
methodologies to “assist in the understanding, modeling, and forward, including Nash stability [14], [15], general metara-
analysis of conflict” [2, p. 2]. tionality (GMR) [4], symmetric metarationality (SMR) [4], and
Strategic conflicts are decision situations in which the com- sequential stability (SEQ) [5]. In the graph model, a state that is
bination of independent choices of two or more decision makers stable for all DMs under a given solution concept constitutes an
(DMs) with different preferences and objectives determines the equilibrium under this solution concept.
outcome [3]. A range of methodologies has been proposed to In the assessment of the stability of a specific state for a
investigate strategic conflicts, including metagame analysis [4], given DM, preference information plays a significant role. Some
conflict analysis [5], graph model for conflict resolution [2], models require that a DM’s preferences be expressed as car-
theory of moves [6] and its fuzzy counterpart theory of fuzzy dinal payoffs, indicating both ranking of states and the degree of
moves [7], [8], and drama theory [9]–[11]. As Kilgour [12] puts preference of one state over another; others accept ordinal pref-
it, an essential feature of these methodologies “is game-theoretic erences (ranking of states from most to least preferred, with ties
roots; all are essentially game theory variants that have been de- allowed). The graph model requires only relative preferences, so
signed to yield better decision advice or more compelling struc- all that is required is that a DM’s preferences between states be
tural insights.” expressed using the binary relations “is (strictly) preferred to,”
It is impossible to discuss the modern development of game , and “is indifferent to,” . In a social conflict context, it is
theory without referring to the landmark monograph entitled usually difficult to obtain DMs’ values and risk attitudes, which
are required to determine cardinal utilities. Because the graph
model needs only relative preference information, it is well-de-
Manuscript received August 6, 2003; revised December 18, 2003. This paper
was recommended by Associate Editor J. M. Tien. signed for modeling and analyzing social conflict. In fact, the
K. W. Li and K. W. Hipel are with the Department of Systems Design Engi- graph model can handle both transitive and intransitive prefer-
neering, University of Waterloo, Waterloo, Ontario N2L 3G1, Canada (e-mail: ence information. In those rare cases when cardinal preferences
w8li@engmail.uwaterloo.ca; kwhipel@uwaterloo.ca).
D. M. Kilgour is with the Department of Mathematics, Wilfrid Laurier Uni- are known, the ordinal rankings they imply can be employed to
versity, Waterloo, Ontario N2L 3C5, Canada. (email: mkilgour@wlu.ca). calibrate a graph model.
L. Fang is with the Department of Mechanical and Industrial Engineering, In some situations, however, information about the DMs’
Ryerson University, Toronto, Ontario M5B 2K3, Canada. (email: lfang@ry-
erson.ca) preferences is incomplete, and/or the conflict under consider-
Digital Object Identifier 10.1109/TSMCA.2004.826282 ation is on-going, so that preferences can be expressed only in
1083-4427/04$20.00 © 2004 IEEE

a key advantage of the graph model is flexibility. our research is different from the incomplete pref- i) is asymmetric. i. if . [18]. i. more information becomes available. [19] introduce a new binary graph model. 4. — all unilateral disimprovements [2].e. tract meaningful equilibrium information based on generalized To characterize a DM’s preferences between states. plete preference structures assuming a vector utility function iii) is symmetric. a set of state . in this re- is true. We acknowledge . suming . when characterization of a DM’s decision possibilities at a given state. hold true at the same time. other. . at any state . Section IV then investigates — all states that DM prefers interrelationships of the solution concepts within each set of sta. Section III de. where .e. The preference structure characterizing DMs’ analysis of a conflict model with incomplete preference infor.508 IEEE TRANSACTIONS ON SYSTEMS. MAN. state in one step.. where a DM can direct — all equally preferred states the conflict from one state to another. this new stability definitions. specifies DM ’s immediate decision possi- where indicates DM ’s preferences. irreversible moves are simply identified by unidi- relation “is uncertain about. to for DM . search. these four subsets constitute a partition of the state set .” . to a triplet . to . where [23] to represent DMs’ cardinal utilities. to express a DM’s uncertain rectional arcs. with describing the possible be partitioned into four subsets as follows: moves among the states controlled by DM . i. . With this new binary relation This research takes a different approach. and cannot erence relation used in game theory and economics [22]. containing all states that can be reached by DM from preference relationships among the states . . VOL. Moreover. indecisiveness of DMs. each node denotes a feasible state. ample. compare the two states and assess which is preferred to the tems for incorporating uncertain preferences. for any . GMR... press unknown or uncertain preference between two states as then exactly one of . from the first state to the second. II. the incomparable relation in the decision-analysis literature erences is considered here. SMR. Built upon the above partition. However. for describing preference uncertainties. — all states that DM prefers Illustrative examples with two-DM and three-DM graph models to. [23]. is founded upon relative preferences and permits DMs to ex- iv) is strongly complete. or well as strict preference and indifference. A typical procedure is to define incom- i. and focus on specifying the minimum requirements for lation “is uncertain” to represent DM ’s uncertainty about preferences in order to ensure that a state is stable. [25].e. NEW PREFERENCE STRUCTURE Because of the strong completeness property of . NO. where . hence. In a reachable from state by DM . Li et al. If DMs have uncertainty about what [26]. and are different states in ): Moreover. DM ’s reachable list from a given comprised of a set of DMs . for ex. and reflect the degree of uncertainty between the known and the means that DM is indifferent between and (or equally true preferences. In a directed graph. As mentioned earlier. in handling irreversible moves.e. bility definitions and across the four sets of stability definitions. preferences in the graph model are ex- Ben-Haim and Hipel [20] develop information-gap models pressed in terms of a pair of binary relations on . — all states that are indifferent ments in Section VI.e. Note that only uncertainty in pref. The paper concludes with some com. and symmetric. JULY 2004 an uncertain fashion. — all states for which DM is uncertain about preference relative to . are presented in Section V. and SEQ in the context of two-DM identify four subsets of the state set : and general -DM conflict models. which of two states is preferred. Within the paradigm of the graph model. may be a convenient tool [24]. uncertain Section II introduces the new preference structure and its preference generally reflects a lack of information. velops four different stability definitions for each of the solution Based on this preference structure.. a pair of binary relations. i. a DM can concepts Nash. our approach . the preference structure in the graph model is extended from the uncertainty in preferences at the beginning. As pointed out by Fang et al. where the completeness of is relaxed to characterize ii) is reflexive. it is assumed that each DM correctly knows the prefer- ences of all DMs in the conflict. The aim is to extend the graph model for preference structure should possess the following properties (as- conflict resolution to handle uncertainty in DMs’ preferences. and a set of directed bilities at state . [27]: if two items seem radically different. using a parameter to where indicates that DM prefers to . but not vice versa. the DM may be able to Based on this extended preference structure and different sys. AND CYBERNETICS—PART A: SYSTEMS AND HUMANS. if then . preference between two states for a two-DM conflict model. including any uncertain pref. and try to ex. [21] consider the robustness prefers and ). preferences is extended here by introducing a new binary re- mation. However. if then . . a set of feasible states . a conflict model is In the graph model. and an arc with orientation — all unilateral improvements from one state to another indicates that DM has a legal move from state for DM . Sakakibara et al. hypergame analysis feel that it does not make sense to compare them and. Note that the uncertain preference relation is different from erences that may be present. can thus graphs . DMs may possible outcomes may arise in a conflict. refuse to offer a comparison between them. 34. from state for DM .. or whether they are indifferent.

if a DM is in a poor position. GMR. with respect to the current state . GMR. and SEQ in two-DM conflict Definition 4: Sequential Stability ( ): For . consider the stability def. Li et al. 3. ciated with the uncertain gain) but. extension of stability definitions for the solution concepts Nash. denoted by III. Li et al. denote the two DMs as ences: only states that leave him or her no better off for sure DM and DM . there uncertainty. i. denoted by . . simply excludes uncertainty in is Nash stable for DM . such that and for all initions for Nash. SMR. After definitions of a unilateral move (UM) and a unilateral improvement-uncertain move (UIUM) by is sequentially stable for DM . models. If any DMs may exhibit different attitudes toward the risk associated countermove leads to a less preferred or equally preferred state with uncertainty in preferences. SMR. there exists Based on this extended structure. he or she considers not only right column of the table will be explained by Theorems 2 and the responses by DM but also his or her own counterresponses. a state is symmetric metarational for DM . a state and evaluates sanctions from his or her opponent. sanctions. denoted by . DM may be tempted to move to a UI or a state with uncertain preference from state . definitions and properties of the aforementioned and other solu. where a sub. and . A state The next extension. Hence. therefore.) uncertain. form b. When DM as- rated. represents stability for only if or . [19] devise four extensions of stability def- exists . and define if and constitute sanctions. and that his or her his or her position. A state is SMR for DM only if the sanction from DM The four systems for extending stability definitions for solu- cannot be avoided by DM ’s counter-response . such that . On the contrary. the most aggressive DMs. there exists at least one naturally extended to general -DM models. tion concepts Nash. denoted extensions of stability definitions in the context of preference by . SMR. the DM may tend to be conservative and resist changing stop after his or her opponent countermoves. Two-DM Models The above extension of definitions (form a) assumes that DM Stability definitions are now separately formulated for the is aggressive when deciding whether to deviate from the cur- four extensions mentioned on the left in Table I. If a DM’s current position is satis- at state . opponent will act to sanction DM ’s move without regard to his and has little to lose. DM will not consider states with uncertain prefer- initions for extension a. when evaluating possible sions a–d is being entertained. For mathematical . such that . Form a. These considerations demonstrate the need for a range of . a group of DMs are introduced. if and preferences when the focal DM considers incentives to leave only if . relative to current state . Table I captures the key features of these extensions.LI et al. but possible countermoves by his or GMR. refer quires the sanctioning responses from DM to be credible. to [2] and [28]. and Corollaries 2 and 3 in Section IV. The sesses his or her individual stability. a state models are presented next. to benefit DM . DM will stay according to circumstances. It is Under Nash stability. if and only if for every . he or she may become aggressive and be or her own benefit. This attitude may also change for DM . either directly (UI) or potentially (unknown preference). willing to accept the risk associated with a potential improve- Definition 3: Symmetric Metarationality ( ): For ment. if and only if for every . First.: PREFERENCE UNCERTAINTY IN THE GRAPH MODEL FOR CONFLICT RESOLUTION 509 TABLE I DIFFERENT EXTENSIONS FOR SOLUTION CONCEPTS WITH PREFERENCE UNCERTAINTY — all states reachable by DM uncertain preference. depending on how preference uncertainty is incorpo- SMR looks ahead one step further than GMR. and SEQ for two-DM conflict models. DM cannot improve his or her position worth noting that this set of definitions is different from the for certain because he or she has no UIs or moves to states with stability definitions without preference uncertainty as discussed . Definition 1: Nash Stability ( ): Let . [18] put forward an at least one .. (The latter may from state for which DM ’s preference relative to is turn out to be UIs when preferences become known. The only difference between GMR and SEQ is that SEQ re- tion concepts without considering preference uncertainty. a state is general metarational for DM . and SEQ for two-DM conflict . rent state (because he or she is willing to accept the risk asso- script connected to a solution concept indicates which of exten. Definition 2: General Metarationality ( ): For . A.e. For simplicity. Note that DM expects that the move sequence will factory. these stability definitions can be if and only if for every . SOLUTION CONCEPTS WITH PREFERENCE UNCERTAINTY . Different her opponent DM are taken into account under GMR.

a) if and . denoted by . NO. Definition 17: UM by : Let and . ence uncertainty is allowed when a DM considers both incen. if and only if for every . -DM models after introducing the definition of UIUM by a Definition 5: Nash Stability ( ): Let . in that all states in only if .. . if and only if for every . is defined is Nash stable for DM . set of all states that can be reached through any legal sequence In the third extension of stability definitions (form c). In the latter case. [2].a attained via sequences of “joint moves” by some or all DMs in state is sequentially stable for DM . there exists ii) if . To be credible under sequential stability. i) if and . i) adds states that are UMs from state for all . denoted by . vidual DMs in the group. A state A UM from by . Let be a nonempty subset of all DMs. there exists at containing consecutive moves by any DM. such that and for all . there exists at bility examines the credibility of the sanction by the opponent least one such that or . a state is symmetric metarational for DM . if and only if for every . When assessing the stability of a state for a given DM. if and only if for every . AND CYBERNETICS—PART A: SYSTEMS AND HUMANS. MAN. Form d represents sta. JULY 2004 by Fang et al. then at least one . then and by . 34. if and only if for every . and . -DM Models Definition 6: General Metarationality ( ): For . the opponents are a group of two or Definition 7: Symmetric Metarationality ( ): For more DMs. or . . so. A state in which the last move is not made by . which means that the sanction by Definition 15: Symmetric Metarationality ( ): For the opponent must benefit (i. denoted by . denoted by . there exists at least one more than once. there exists relative to ). a DM may move if and only if for every . then Definition 11: Symmetric Metarationality ( ): For and .510 IEEE TRANSACTIONS ON SYSTEMS. denoted -DM models. denoted by . then any state can bility for the most conservative DMs.a their responses are legal sequences of UMs by members of this state is sequentially stable for DM . if and regarded as the reachable list of . from deviating from the original state. the opponent is a single DM. sequential sta- . if DM to leave the current state but is sufficient to deter the DM there are two or more DMs who can make a move from to . a state is symmetric metarational for DM . This is achieved by least one such that or . DM . distinguishing from : if The last extension of stability definitions (form d) adopts the there is only one DM in who can move to .e. A UM by a Definition 8: Sequential Stability ( ): For . there . of UMs from state by some or all DMs in . let be the set of all last DMs in legal sequences from tives to leave a state and sanctions to deter him or her from doing to . A state group of DMs. If . such that or . if and only if for every .e. Be- for all . if and inductively by: only if . be added to because there exists a sequence from to Definition 13: Nash Stability ( ): Let . denoted b) if . Definition 16: Sequential Stability ( ): For . To extend the stability definitions in Section III-A to . but only because previous definitions cannot Below these stability definitions are extended to general analyze conflict models with uncertainty in preferences. by all DMs in . exists such that or and or To determine . . it a state is general metarational for DM . In a legal sequence. -DM model with . a state is general metarational for DM . can be achieved by some or all DMs in without participation Definition 14: General Metarationality ( ): For of any DM in . if and only if for every . Let denote the such that . but not consecutively. a state is general metarational for DM . This feature is retained in general -DM conflict such that or and or models: sanctions must be credible for sequential stability.e. is a UI for from to . denoted by is necessary to examine possible responses by other DMs. . 4. Let denote the set of states that . The set can be is Nash stable for DM . a state group of DMs is defined by a legal sequence of UMs by indi- is sequentially stable for DM . if and only if for every . tain UM for the mover. any individual . prefer. denoted by Furthermore. . the definition of countermoves by a group must by . . is Nash stable for DM . B. denoted at least is a move from to a state of uncertain preference for by . there exists be introduced first. there exists at a two-DM model. is a member of if and only if . i. cause the opponents are a subset of DMs instead of a single DM.. a state is symmetric metarational for DM . if and only if . recall that in the two-DM case. i. it is necessary to screen out sequences . a state view that uncertainty in preferences is not enough to motivate a . In . there exists at least move in the sanction sequence must be either a UI or an uncer- one . denoted by subset. denoted by . such that or . while ii) adds those other states that can be Definition 12: Sequential Stability ( ): For . VOL. a member of .. Definition 9: Nash Stability ( ): Let . while in an least one such that . then and Definition 10: General Metarationality ( ): For .

if and only if for every . simply an extension of . if and for all . and SEQ can be extended naturally to the -DM situation. denoted by to . such that or . such that or . GMR. A state cepts of the graph model. a state is general metarational for DM . SMR. and . If . at least one . a state is general metarational for DM . [28] explore the interrelationships of solution con- Definition 23: Nash Stability ( ): Let . denoted by . and . if and only if for every . if and only if for every . such that . the definition of the set of is Nash stable for DM . INTERRELATIONSHIPS OF SOLUTION CONCEPTS WITH state is sequentially stable for DM . denoted in Table I. and SEQ still hold true. there exists at least or potential UIs for a group of DMs. by .LI et al. A UIUM from by . a state is symmetric metarational for DM . a state is symmetric metarational for DM . Note that is . and . if and only if for every . a natural question is whether the interrelationships Definition 24: General Metarationality ( ): For among Nash. a state is symmetric metarational for DM . denoted by . GMR. . such that . such that and for all -DM models in the corresponding extensions of definitions. there exists at SMR. Definition 27: Nash Stability ( ): Let . A state such that or and or is Nash stable for DM . at least one . the sets and represent the responses or cred. and is Definition 29: Symmetric Metarationality ( ): For defined inductively by: . least one . where includes all Definition 26: Sequential Stability ( ): For . if and only if for every .: PREFERENCE UNCERTAINTY IN THE GRAPH MODEL FOR CONFLICT RESOLUTION 511 can result from any legal sequence of UIs and/or uncertain UMs exists . there exists one . by . or for all . For the same reason. denoted by . . denoted by PREFERENCE UNCERTAINTY . and SEQ exist for both two-DM and -DM least one . there exists at GMR. denoted by then and . conflict models within each of the four extensions of stability Definition 25: Symmetric Metarationality ( ): For definitions. denote by the set of all last Definition 28: General Metarationality ( ): For DMs in legal sequences of UIs and/or uncertain UMs from . respectively. denoted in forms b and d are identical. . only if . Now that four graph-model solution is Nash stable for DM . such that and for all starting from by some or all DMs in . only if . the labels a–d refer to the four ways of Definition 33: Symmetric Metarationality ( ): For incorporating uncertain preferences as shown in the left column .a a) if and . Definition 22: Sequential Stability ( ): For .a Definition 20: General Metarationality ( ): For state is sequentially stable for DM . if and only if for every . denoted by . denoted by demonstrates that the same interrelationships among Nash. a state is symmetric metarational for DM . Since Nash stability does not take into account countermoves Definition 21: Symmetric Metarationality ( ): For by opponent(s). denoted by a single DM .a IV. such that or . For any and . a state is general metarational for DM . such that or . . if and only if for every . there there is no difference for Nash stability between two-DM and exists . . can now be developed. such that . denoted by . if and only if for every . such that . there exists Definition 19: Nash Stability ( ): Let . a state is general metarational for DM .a UIs or potential UIs (with uncertain preferences) from state by state is sequentially stable for DM . . degenerates to . there exists at least one . preferences. Theorem 1 . . Fang et al. only one DM . denoted Theorem 1: Interrelationships of solution concepts within by . Definition 34: Sequential Stability ( ): For . if and In general -DM conflict models with preference uncertainty. ii) if . if and only if for every . A state is Nash stable for DM . state is sequentially stable for DM . Stability definitions for Nash. denoted by . If this group consists of one . there exists at least . if and only if for every . then Definition 30: Sequential Stability ( ): For . As in the two-DM case. there exists Definition 18: UIUM by : Let and . Definition 31: Nash Stability ( ): Let . is a member of . while contains all combinations of UIs . if and only if for every . then and least one . there i) if and . only if . Definition 32: General Metarationality ( ): For ible sanctions by DM ’s opponents against ’s move to . then and exists such that or and . if and UIUM by a group of DMs . SMR. . A state Based on this understanding. if and concepts have been extended to accommodate uncertainty in only if . there extensions. denoted by are of particular interest. denoted by . . there exists at b) if . Nash stability definitions in forms a and c. if and only if for every .

Extensions (GMR. in Fig. Theorem 1 implies that the possible responses from the opponent(s). Proof: The proof is given for the two-DM case under Ex- tension a. state with uncertain preference relative to his or her current state . First. orem 2 and Corollary 2 are illustrated by the Venn diagram in These interrelationships are illustrated in the Venn diagram Fig. though . but . and SEQ. Corollary sets under different solution concepts and/or different forms of 2 gives this result. state . . b–d. respectively. solution concepts. SMR. according to definition 3. the focal DM regards as a valid sanction any at least one such that . GMR.. four different extensions of stability definitions solutions concepts. 34. a DM who follows the stability definitions in exten. . it is not necessary that risk associated with the preference uncertainty. . . while extension c implies that a DM is ag- . Discarding the information about . for every . For any . 1. . . Therefore. but when assessing a response from his Because . For any . and hence form a and same interrelationships exist for group stability. a DM who takes the stability definitions in extension d demonstrates a conservative nature. JULY 2004 . Hence. Because sponse from his or her opponents is not deemed to be enough . therefore. First. while a re- to the above argument. AND CYBERNETICS—PART A: SYSTEMS AND HUMANS. and SEQ stability based tensions discussed in Section III. for every . In this case. This result is stated in Corollary 1 Corollary 2: . For . but aggres- response from state makes DM better off or sive with sanctions. according to definition 2. so. then . as long as DM ’s counter. . if state . stability from a single DM’s viewpoint in terms of different . Theorems 2 and 3 capture the interrelationships of under . the same interrelationships of Nash stability for any can be introduced in the same way to denote equilibria individual DM hold true for Nash equilibria as well. Theorem 2: Interrelationships of Solution Concepts across Theorem 1 examines the interrelationships of individual Extensions (Nash). state (uncertainty is allowed at the sanction end). Since individual stability. two cases may arise. according to the above argument. the risk associated with the uncertain preferences (uncertainty . we prove that if . This intuitive interpretation of the four extensions of sta- (for two-DM case) or the response by DMs in belongs bility definitions is helpful in understanding the fourth column to (for -DM case). Theorem 3 depicts the interrelationships of the other three In Section III. NO. . GMR. and . Hence. for gressive at the incentive end. However. . . Consistent with the set notation for Next. b–d. then . as shown in Table I). are satisfied because no belongs to . there exists or her opponents. and form b and form d are exactly the same. . trary. Theorem 3: Interrelationships of Solution Concepts across Intuitively. and Corollary 1: . across the four ex- are developed for Nash. then . . SMR. and Fig. and hence. according is allowed at the incentive end as shown in Table I). extensions. not deviate from the current state unless he or she is guaranteed . . it is easy allowed at the sanction end. . 2. The same interrelationships hold true for the other three extensions. . it is obvious that set . . argument can be easily applied to the other extensions. the equi. MAN. as well. libria of a conflict model. . but is not credible for DM or under the solutions concepts across the four extensions of definitions . VOL. and Similarly. extension b in- or . if the response by DM belongs to end. The definitions in .e. Second. . . Hence. again. but conservative at the sanction . because the focal DM will If state . This proves that . In this case. a better state (uncertainty is not allowed at the incentive end. which implies that both includes all Nash equilibria under extension a. Second. . 4. since . i. and solution concept if it is stable for all DMs in the model in terms and is obvious since Nash stability does not consider of the same solution concept. . and . . On the con- to see that definition 2 is satisfied. but it can be easily carried out for the other three ex- tensions and the -DM situation in a similar fashion. as shown in Table I). A similar formally. Similarly. which implies defi- nitions 2–4. In this case. it can still block DM of Table I. sets and . Recall that a state is an equilibrium under a Proof: The equivalence of and .512 IEEE TRANSACTIONS ON SYSTEMS. . Similarly. dicates that a DM is conservative with incentives. extensions b and c display a DM’s mixed attitudes toward the Generally. form c. . and sion a tends to be aggressive in that he or she is willing to accept If state . SMR. . Interrelationships of solution concepts within each extension. 1. b–d. If state . on different ways to take into account preference uncertainty. two cases may arise. The same interrelationships hold true for the The interrelationships of Nash stability characterized by The- other three extensions. to deter the focal DM from leaving the current state unless the there exists such that and for response leaves the DM no better off for sure (uncertainty is not all . and SEQ).

If Next. Now. If . we prove that . . then least one such that . Again. such . The . It is trivial to extend the proof to the for all . then . then for every and . . and hence . for DM under Extension a. Therefore. so each . . then for every . Thus. Because . are discussed here. If . any UI from state for DM can and according to Theorem 1. there exists . satisfies or . . there exists at least one such that . there exists at SMR: Assume . according to Definition form d. and . in the following argument. Case 2: . there sanction of extension d. which implies that or . If . there exists at least one such that . exists such that and for all SEQ: If . The proof proceeds by sequentially considering each of the and . hence. 2 for every . in GMR and SMR. and subsequently. in SEQ. and stable for DM .LI et al. . Definition 1. Since . is also stable for DM in terms of . Otherwise. If . for such an . because . and . Now. . such that . and . . hence. there following proof is developed for each of the three solution con. such an . it follows from definition 3 that for every aforementioned . there exists at least one For Extension d. hence. SMR and SEQ. such an satisfies or and. it is natural to have lution concept is applied when asserting a state is stable. GMR: If state . If . such that . or . Because . such that and cepts for two-DM models. and with SEQ: Again. So. For each so . The and for all . then for any . and The following arguments assume that d. if . if . . 6 shows that for every . according to definition 3 for every . . and stable and. According to definition Fig. and . . which cepts. and automatically . which makes any In the following. If state . as defined in definition 15. or and or for all . we prove that . such that . and hence as per Theorem 1. and that implies or and implies as per Theorem 2. and For Extension d. Therefore. exists at least one . there exists at least one GMR: If . then that and for all . Case 1: . for each three solution concepts. Because state counter-response of DM futile as well under the weaker . assume . . there exists . Following the same argument as for extension c. the proof is trivial if . . . according to For Extension b. we prove that stability of form b implies stability of For Extension c. because . then for every . . extensions.: PREFERENCE UNCERTAINTY IN THE GRAPH MODEL FOR CONFLICT RESOLUTION 513 For extension b. then For extension d. According to Theorem 1. . implies that state as per Definition 11. Because . then Definition . and . Such an satisfies same interrelationships hold true for the other two solution con. Under this circumstance. Because . 2. . First we show that . if state is stable in terms of GMR. i. is . SMR and SEQ argument. . state . then such that . there exists at least that . then for every . This proves that state . Again. one such that . SMR. Next. For extension c. we have since . according to the above Firstly. and . such that . Following the same argument as above. and hence . consider . hence . .. . Proof: Since the three solution concepts. for each . c. according to Theorem 1. if . it is necessary to identify which so. GMR. hence . because . and . . there exists such SMR: Similarly. we -DM case by replacing the countermove with can assert that . there exists . . state . there exists . be blocked by a move by DM to a state . which implies that . notice . or . hence . . there exists Under extension c. Interrelationships of stable states for Nash stability across the four 7. . there exists such that and for all . if and . for each . we assume that . and as per Theorem 2. Hence. for every . which proves that . If . if . then . such that and . If SEQ.e. for each . the corresponding solution concept under Extensions b. it follows that because .

However. . Hence.514 IEEE TRANSACTIONS ON SYSTEMS. if and . All these transitions are reversible. But . . 3. Therefore.e. transitivity of the strict preference relation is assumed for the Otherwise. which implies hold true for equilibria as indicated in Corollary 3. and the only . so for all . so state under Definition 5. least one such that or . and state . The . state is tions . we prove that if state is stable for DM under exten- sion c. i. then . state is GMR for under Exten- column ( ). SMR. there exists . within each extension as described in Theorem 1. respectively. and Corollary 3: . transi. . But any DM. JULY 2004 TABLE II PAYOFF TABLE FOR A 2 2 2 GAME WITH UNCERTAIN PREFERENCES . MAN. . where the four entries for . and SEQ for DM under Extension b may that is stable under extension c may not be stable under Exten- not be stable for DM under extension a or c. denoted by row ( ) and Under the assumptions. 4. .and . reading from left to and or for all . then either or . Graph model for a 2 2 2 game. there exists such that or . Note that the ma- exists at least one such that or and trix form of a 2 2 game model with ordinal payoffs generally or for all .e. SMR and SEQ. and verse of the second part of Theorems 2 and 3 does not hold true. represent the preferences for states to . . Similarly. . to .. . is for under this definition. However. Assume . The even and for . NO. Hence. Consequently. Either or . and and . constitute a sanction for according to Definition 2 or 6. As pointed out in the first part of Theorem 2. but here assume and . The preferences are given by and for DMs and . . where the directed graphs indicate the available moves. According to Definition 10 (GMR under Extension first and second entries inside the brackets in each cell stand for c). then Fig. SMR. Notice that . in that state does not same column. so if . implies the transitivity of the strict preference relation : for SEQ: .and . in general. so if . the interrelationships of solution concepts across . AND CYBERNETICS—PART A: SYSTEMS AND HUMANS. then is also stable for DM under extension d. Hence. Each cell in the matrix of Table II represents a sion c because the only UI by from state will leave state. the . so if . . because and . transitions . a state Nash. But . trols the transitions across columns in the same row. Therefore. for solution concepts GMR. In addition.. Otherwise. However. i. . state. the response from by DM constitutes a credible sanction for DM . GMR: Assume . there exists at subsequent examples in this paper. GMR. re- there exists at least one such that or spectively. Hence. Fig. 34. Example 1 demonstrates that a state that is stable in terms of Similarly. shown in Table II. for which a that . in the top left cell in Table II. Example 1: Consider a 2 2 game with payoffs as shown in Example 2: Consider again the 2 2 game with payoffs as Table II. and SEQ. and . controls the transitions across rows in the ever. If we further assume that and . Ac- at least one such that or .. . there higher number means a more preferred state. Otherwise. GMR. and the four extensions of definitions described in Theorem 3 also in that . as illustrated by example 2. or . so for any . leaves better off compared the other two solution concepts. Next. and . but cannot be .e. . so for any . In the graph. The same interrelationships hold true for countermove by from . then . to the initial state . VOL.. This model consists of two DMs.. and . cording to the interrelationship of Nash. and because . there exists such that or . How- model terminology. i. SMR: If . then either or . is not or for . for any . and the situation where both and select their first strategy. then . For instance. equivalent graph-model representation of this game is shown in However. that . state is at state after makes a countermove from state . Hence. for all . for any . and SEQ . Note right. sions a or b. SMR. Therefore. the con. and . 3. and hence. while con. will have no incentive to deviate from his or her current the relative preferences of and . according to Theorem 1.

the DEV’s preferences con- . . by deviating from the current state. or stable. and hence also stable for . but it is not . . these two games are aggregated into one gen- V. ENV can be proactive (P) or reactive (R) in The stability analysis results based on the four extensions of monitoring developers’ activities and their impacts on the envi. Fig. and stable for a DM. SMR. if we further assume that and . and consequently cannot be . no inclu- sion relation holds true for the other three solution concepts. GMR. but not under ex- tensions a–c. or stable. in that . and before ENV reveals its plans in enforcing the relevant environ- SEQ. An illustration is found in the some developers may be more environmentally responsible than second half of Example 3. but no inclusion relationship exists. under extension d. 4. However. Table III gives the matrix form of 3. SMR. Generally. and Theorem 1 guarantees that and . stability definitions for Nash. erences are contingent upon ENV’s enforcement measures. and SEQ. certainty and . in Table III. It can be shown that state is also . except for Nash stability where . where a check ( ) under 1 or 2 for a given a low priority (L). and SEQ across the Example 3: Consider again the 2 2 game with payoffs as four extensions. TABLE III Under these assumptions. and SEQ are dis- ronment. and for . as shown in the first part of Example 3. Consequently. or for provided that and hold true since and . some preference information is only available as uncertain re- Example 4: A two-DM conflict model with preference un. accurate information re- Hipel [1] considers a sustainable development game to model garding DEV’s attitude toward the environment becomes avail- a conflict between environmental agencies and developers. using a Venn diagram. and SEQ in Theorem 1. two-DM and more than two-DM conflict models. However. SMR. . APPLICATIONS eral model with uncertainty in DEV’s preferences.: PREFERENCE UNCERTAINTY IN THE GRAPH MODEL FOR CONFLICT RESOLUTION 515 . is not or stable. . as shown in the last paragraph. .LI et al. In some cases. in terms of GMR. ENV’s preferences tensions of the four solution concepts can be used to analyze are complete and do not contain any uncertainties. In other cases. from ENV’s viewpoint. SMR. the two models. and . . state cannot be or TABLE IV for . state means this state is stable for ENV or DEV under the cor- . From DEV’s own viewpoint. GMR. These observations motivated the introduc- under Extension c. is . . or stable for according to The- orem 1. and for . stable states under Extension b may overlap with those mental regulations. Interrelationships of stable states for GMR. for GMR. it can be argued that this DM is uncertain about its preferences These discussions demonstrate that. Table IV. lations: . would be better off at . . its pref- . . a state is . as shown in Two applications are now developed to illustrate how the ex. . state is . If. . and for . and . . But state is not for according to Definition 10 in that and . Taking advantage of the uncertain preference relation intro- duced earlier. So. and SUSTAINABLE DEVELOPMENT GAMES WITH DIFFERENT DEVELOPERS. In the general model in Table IV. According to the interrelationships of GMR. a state is . SUSTAINABLE DEVELOPMENT GAME WITH ENVIRONMENTALLY As for the number of stable states under Extensions b and UNCERTAIN DEVELOPERS c. and SEQ. then is stable for because . but not . state is tain some uncertainties. or potentially better off at . where state is stable for others. For DEV. Example 3 demonstrates that a state may be stable. (A) HIGH ENVIRONMENTAL PRIORITY (B) LOW ENVIRONMENTAL PRIORITY stable for based on definitions 14–16. SMR. . . DEV may make the environment a high priority (H) or played in Table V. However. Velopers (DEV). after further study. preferences for ENV are uncontro. because . . This able. and stable for a DM. but not versial. preferences for DEV are not so simple. among others as shown in Theorem possible nature and stances. . and . SMR. and hence. so or for as shown above. this model may be reduced to one of the two games shown model has two DMs: ENVironmental agencies (ENV) and DE. . Fig. shown in Table II with assumptions . 4 tion of two different 2 2 game models to characterize DEV’s expresses this relationship.

Later on. AND CYBERNETICS—PART A: SYSTEMS AND HUMANS. speculating on the possibility of improvement. SMR. SMR. responding solution concept. who cited poten- high priority game. they may jump back and forth between the two strategies. they would not consider switching their current strategy and would find all states Nash stable.516 IEEE TRANSACTIONS ON SYSTEMS. Incorporated. The history of this conflict dates each extension of definitions. 4. [29] and Hipel et al. GMR. when a Newfoundland company. The only exception is that. the Fed- a UI from to . More specifically. none of the four states is Nash stable for DEV under a or c. 1996. [30] analyze the Lake Gis- compared horizontally. borne. it will be an equilibrium under all four solution concepts. states are all Nash stable under b and d because no matter which However. In late March 2001. back to June 1995. the story does not end here. JULY 2004 TABLE V TABLE VI STABILITY ANALYSIS RESULTS FOR SUSTAINABLE DEVELOPMENT GAMES. DMS AND THEIR OPTIONS IN THE GISBORNE CONFLICT MODEL (A) UNCERTAIN CASE (B) INDUCED GAMES TABLE VII FEASIBLE STATES FOR THE GISBORNE CONFLICT MODEL under uncertainty DEV can be motivated to deviate from his or her current state. does not contain any uncertainties in this model. which in turn are also GMR. 34. and is indicated by a check in column E. the four sta. because ENV’s preference information Taking into account potential economic benefits to the province. two and only two states are Nash stable. which is located near the south coast of Newfoundland. However. Due to this pressure. the stability the government of Newfoundland and Labrador approved this results for ENV for the uncertain model and the induced games project on December 5. duced its own policy to prohibit bulk water export from major Because Nash stability looks only one step ahead. no UI is available for DEV. The results con. This move triggered fervent criti- are essentially the same. the new premier of New- . state is the status quo. NO. and has a sim- solution concept. proposed a project to export water from Lake Gis- firm Theorems 1–3. for the cism from many environmental lobby groups. forcing Canada Wet and c). Due to the uncertainty in DEV’s preferences. For the two induced strictly ordinal games. Canada Wet sions of definitions for each solution concept. if developers are conservative and not willing to accept the risk associated with preference uncer- tainty. In reality. for any DM [2]. the four to abandon its development in Lake Gisborne. across the four exten. VOL. and SEQ. If one state is Nash stable for both DMs. as predicted under b and d. state is SEQ for ENV for the eral Government of Canada joined in the opposition and intro- uncertain case and low-priority game. or vertically. However. MAN. state is not SEQ for ENV because DEV tially devastating consequences to the environment in the basin does not have a credible sanction to prevent ENV from making and to the culture of the local First Nations. the provincial bility definitions degenerate to two: either examining UIs only government introduced a new bill to ban bulk water export from (Extensions b and d) or UIs plus “potential” UIs (extensions a Newfoundland and Labrador in late 1999. and SEQ. Table V provides some insights into the four forms of stability Example 5: An -DM conflict model with preference uncer- definitions with preference uncertainty for the solution concepts tainty Nash. if motivated by the urgent need for cash. A state that is stable for both DMs Table V(b) indicates that state is such a strong equilibrium if under a solution concept constitutes an equilibrium under this developers give environment a high priority. The stability of a state can be Fang et al. across the four solution concepts within borne water export conflict. drainage basins in Canada. ilar status if they give environment a low priority. and the associated corollaries. But if developers are aggressive and ready to take risks.

shown in Table IX. in general. (A) ECONOMICS-ORIENTED PROVINCIAL GOVERNMENT (B) ENVIRONMENT-ORIENTED PROVINCIAL GOVERNMENT foundland and Labrador. Roger Grimes. Extension a produces are given in Table VIII. 5. and 8 . where the nodes represent feasible states. This integrated graph clearly identifies the moves and Tables X and XI again confirm the interrelationships of the countermoves among the three DMs starting from any state. TABLE VIII TABLE IX PREFERENCE RANKINGS FOR THE GISBORNE CONFLICT MODEL: PREFERENCE RANKINGS FOR THE GISBORNE CONFLICT MODEL: INDUCED UNCERTAIN CASE CASES. four solution concepts from within and across each extension The preference rankings over feasible states for all three DMs of definitions. . for DM Provincial in Table VIII. 5 shows the integrated graph of the Gisborne conflict in Table XI. is transitive. This feature makes it natural to introduce tion. a state in one set of brackets no inclusion applies.: PREFERENCE UNCERTAINTY IN THE GRAPH MODEL FOR CONFLICT RESOLUTION 517 Fig.LI et al. For instance. and extension d the most. this conflict model while DM Support includes not only Canada Wet but other may induce the two cases with complete preference information groups interested in promoting bulk water exports from Canada. the stability-analysis results for the un- the DM controlling it and an “N” means not. DM As pointed out by Fang et al. Integrated graph for the Gisborne conflict model. and hence can handle intransitive prefer- graph model in option form as shown in Table VI. where Applying the solution concepts for -DM conflict models de- a “Y” against an option indicates that this option is selected by veloped in Section III-B. Each DM controls one option as explained in Table VI. DMs 1–3 refers to Federal. As in Table V. In the preference ranking overlap of stable states between extensions b and c. . nomics-oriented or environment-oriented. the low revenue likely generated by water exports was a key consideration. producing the results displayed Fig. an “E” in- arcs along with the DMs specify the moves that can be made by dicates an equilibrium. as stated in Theorems 1–3. each DM. ysis for the two induced cases. In this model. There is some preferences is built into the model. the preference struc- uncertainty in the preference of the provincial government in a ture presented in this paper does not require the transitivity of model of this conflict. The status quo certain case are shown in Table X. The uncertainty in DM Provincial’s the fewest stable states. Among many rea- sons for this decision. Note that this representation considering whether to lift the ban to allow bulk water exports of preference information implies that the strict preference rela- from Newfoundland. This conflict is represented as a three-DM preference relations. and are listed in Table VII. Depending on whether a provincial government is eco- and DM Federal government includes most opposition groups. initiated a review of the Gisborne project. but the turbulence was eased by the release of Justice Minister Kelvin Parsons’s report on October 18. In Tables X and XI. This initiative was widely criticized. [30]. ences. the Provincial prefers state to 7 . . and directed Provincial. . the meaning of DM Provincial is obvious. but as usual. It is still states . . promising that no legis- lation to lift the ban on bulk water exports would be introduced during the coming session of the legislature. is uncertain to all states in any other brackets. . and Support. However. respectively. 2001. [29] and Hipel et al. The decision support system state is the third column of Table VI. model. All eight mathematically GMCR II [31]–[34] was employed to conduct a stability anal- possible states are feasible. but and 7 are uncertain to provincial government is a critical DM in this conflict.

If the support group succeeds. if it stays at a more environmentally conscious position. before the other solution concepts are extended. the temptation of potential economic benefits makes it restless. the interrelationships among solu- tion concepts considered in Section IV may be expanded ac- cordingly. where intransitive uncertain preferences are present. equal. the opposition pressure makes it retreat. four distinct extensions of the stability definitions are developed. Other solution concepts. Except under extension a. STABILITY-ANALYSIS RESULTS FOR THE GISBORNE CONFLICT Illustrative examples demonstrate how the new definitions can MODEL: INDUCED CASES be applied to analyze conflicts and to obtain new insights into the problem under study. There will be no resolution until the key DM. do not make sense un- less preferences are transitive. Instead. states . Depending on how the uncertain preference information is utilized. and unknown or uncertain preferences in comparisons of two states.518 IEEE TRANSACTIONS ON SYSTEMS. . By comparing the stability and equilibrium results for the un- ACKNOWLEDGMENT certain case and the two induced cases. This paper only redefines four of the graph-model solution concepts in the context of the new preference structure. These four solution concepts were chosen because they are the easiest to work with and they do not require the assumption of transi- tivity of preferences. and state when the MODEL: UNCERTAIN CASE provincial government is environment-oriented. This critical feature predicted by Extension a implies that this conflict will probably remain unsettled into the future. The authors are grateful for the constructive comments from . SMR. the conflict is likely to settle at state . NO. 4. the implications of assuming transitivity for uncertain preferences should be ad- dressed. The stability-analysis results obtained under extension a de- pict the dilemma currently faced by the provincial government: when it allows water exports. clarifies its stance. and SEQ. to accommodate uncertainty in DMs’ preferences in two-DM and -DM conflict models. as is the case in example 4 in Sec- tion V. Dr. Interrelationships of the four solution concepts within each TABLE XI extension and across the four extensions are then investigated. After the new preference structure is incorporated into the other solution concepts. some insights into the conflict can be garnered. AND CYBERNETICS—PART A: SYSTEMS AND HUMANS. This informa- tion is significant to both the opposition group and the support group: they are put in an open arena to compete for the favor of the provincial government. GMR. MAN. and 8 are equilibria (under some or all four solution con. JULY 2004 TABLE X that states and 8 are potential resolutions when the provin- STABILITY-ANALYSIS RESULTS FOR THE GISBORNE CONFLICT cial government is economics-oriented. VI. if the opposition (in our model. The uncertain preference rela- tion may not be transitive. which means that the provincial government will lift the ban for water export. the resolution tends to be either state or 8 . It is not simply a coincidence tainty can be considered in stability definitions. VOL. This unique preference structure makes it pos- sible to redefine four graph-model solution concepts. which corresponds to the scenario that the provincial government will maintain its ban over the water export issue. so. 34. such as limited move stability and nonmyopic stability. the federal government is the representative of this group) effectively influences the provin- cial government. L. Hamouda regarding the ways in which preference uncer- cepts) in the uncertain model. the provincial govern- ment. CONCLUSION AND FUTURE WORK A new preference structure is introduced to handle strict. Nash.

Math. D. Fraser. Li. Nakase. government agencies. [7] A. 1998..” IEEE Trans.. the U. W. Confrontation Analysis: How to Win Operations Other than inventory management.” IEEE Trans. South Africa. pp. Award. Behavior. ship. W. as well as Man. Fang. 1951.A. M. 33.S. ematics. and L. UK: Cambridge Univ. D. Feb. Germany. Germany. “A decision support system for conflict resolution. Centre National de la Recherche Scientifique (CNRS) Research Fellow- [17] M. Prog. NJ: Pren.” Eur. 207–223. 2002. and a [9] N. [4] N. Hipel. formal techniques to problems in international security and arms control.Sc. MAN AND CYBERNETICS . 2–3.. J. pp. W. 7. international security. Wang. Fraser and K. B. and logistics and supply-chain management. Fuzzy Syst. Ben-Haim and K.D. 1989. An Introduction to Game Theory. Wilfrid Laurier Group Decision and Negotiation. 1953. Xiamen. Nat. G. W. Press. Group Decision and Negotiation.Sc. 2001. Design Eng.D. [34] K. Man Cybern. von Neumann and O. Hipel. G. “Book review: Theory of moves. 1995. 2003. Conf. conference and encyclopedia articles. degree in control sci- ical Behavior. of Systems Design Engineering at the University of Waterloo. 1980. Symp. Boggess Award from [23] E.” Eur.K.” in Proc. Fang. in 2003. Cybern. vol. Wang. Kilgour. vol. and coalition formation. from Xiamen University. the University of Toronto. in 1970. Cambridge. Morgenstern. vol. vol. pp. Studies Rev. [13] J. M. no. W. The Graph Model for Conflict Resolution. K. Conf. fair division. N. pp. vol. 2002. vol. and 1975. environmental management. Canada.. Fang. 1984. Water Resources Environ.” agement Sciences and of numerous societies across a range of disciplines in- IEEE Trans. the University of Waterloo. Hipel. vol. Automa- tion. and the Ph. M. L. M. system GMCR in environmental conflict management. Bennett. vol. and L.S. Oper. Hipel. envi- [20] Y. 3rd Int. His primary [30] K. Fang. civil engineering from the University of Waterloo. M. “Gisborne water export conflict at Wilfrid Laurier University. ture: Decision Models and Applications. and X. Cybern.” in Proc. 2002. Fang. Econ. pp. 48–49. 14–23. New York: Oxford ship. 3 ed. degree in mathematics from [26] W. Design Engineering. 1962. M. Monbusho Kyoto University Visiting Professor Posi- [16] P. Ordinal Information and Preference Struc. in 1968. pp.. 1994. Univ. respectively. Paradoxes of Rationality: Theory of Metagames and Polit. Canada. Canada. “2 2 games with firms. degree in civil engineering. M. France.Sc. 32. 13–18. 1993. He has also pioneered the development of computer 1999. He is currently a to the game of chicken. vol. [31] X. pp. Kilgour. C. Press. sertation. respectively. pp. Cambridge. Okada. “The graph model for conflict resolu. [11] P. Department of Automation (previously. Vincke. hydrology.” IEEE Trans. Peng. Fang. (IWRM). 1988. Aumann. no. [10] . F. 608–623. systems for decision support in strategic conflict.” Fuzzy Sets Syst. Bennett. Kilgour is a Member of the Institute for Operations Research and Man- system for interactive decision making — Part I: Model formulation.. Math. Springer-Verlag. 42–55. [1] K. with research and teaching leaves in [28] L. Hipel. and he has applied game theory. University of Waterloo. the M. Kevin W. 41.” Future. Canada. 1. 2002. respectively. A. the IEEE TRANSACTIONS ON SYSTEMS. “A decision support system for interactive decision making — Part II: Analysis and output interpretation. and D. olution. the Norbert Wiener and Outstanding Contribution Awards from the IEEE metrica. Manage. Li. J. Man. 1950. Com. L. nine edited books. pp. (ICWRER). NJ: Princeton Univ. 2002. and 1973. “Non-cooperative games. K. Roubens and P. 3. 2–3. pp.Sc. 109. 83.” Annals Math. 12. Dept. “Modeling decision in international relations: Game theory and beyond. Brams. Interactive Decision Making: Computat. 429–449. . Sakakibara.LI et al.A. G. pp. Kilgour. K. 1995. vol. MA: MIT Press. degree in Negotiation. “The application of robustness tion (AWRA). War. 104. 465–482.Sc. Switzerland. 283–284.. [14] J. Tunisia. Sci. 2003. tion. 1971. 1. Bennett. Director of the Laurier study. K. vol. “Conflict resolution: Theme overview paper in conflict res. Theory of Moves. Japan. in 1991 and tions. G. 99. 11. Oxford. IEEE Int. K. “Conflict models in graph form: Solution concepts and their interrelationships.R. vol. China.. W. M. Canada. 2002. since 1973.” Part-Time Instructor in the School of Business and Economics. Waterloo. DC: C4ISR Cooperative Res. held a Canada Council Killam Research Fellowship. “Drama theory and its relationship to game theory: I and II. Res. 3. He is Professor of Systems Design Engineering at Sci. “A decision support Prof. Ed. W.” Mershon Int. W. Hipel. Oper.: PREFERENCE UNCERTAINTY IN THE GRAPH MODEL FOR CONFLICT RESOLUTION 519 REFERENCES [33] . 80–100. 2002. these decision technologies are water resources management. Waterloo. D. 39. W. ence and the M. He is an Associate Editor of many international journals including analysis to the conflict with incomplete information. J. Canada. His research interests include decision analysis. C. J. 2003. Ok. Conflict Analysis: Models and Resolu. “GMCR in negotiations. Germany: at Wilfrid Laurier University. 1992. The main application areas of mamet. Karray. Kilgour. Systems. 33. Ham. 432–436. Kress. 1970. vol. 151–156. J. 33.. Integrated social science. pp. Zhang.. 1985. Syst. Preference Modeling. Systems [8] K. negotiation and arbitration. the W. Peng.. Kandel and Y. and Japan Society for Promotion of Science (JSPS) Fellowship. and the University of Waterloo Award for Excellence in Research. Waterloo. decision theory. 1994. W. and environmental management. U. Berlin. New York: North-Holland. 319–340. Toronto. 455–462. Englewood Cliffs. Waterloo. Dept.A. Conf. tion with information-gap uncertainty in preferences. vol. His major research interests are the devel- [19] . pp. Hipel. 2003. Peng.” in Proc. Theory. “Stability definitions for 2-player conflict models with uncertain opment and application of conflict resolution and time series analysis techniques preferences. vol. Feb. Stellenbosch. C. Prof. pp. Contr. M. Nash. and the Ph.. W. Schmitz..A. [24] P. Syst. 286–295. Washington. pp.. 9. 1998. 2.” in Proc. Marc Kilgour received the B. Fang. “Utility theory without the completeness axiom. Dresden. Syst. Cook and M. [2] L. no. Kilgour. W. 145 journal papers. W. [3] D.” Ph. M. Howard. and vol. Post-Doctoral Fellow in the Department of Systems Aug. and related Water Res. Acad. “Equilibrium points in n-person games.” in Encyclopedia of Life Support Systems (EOLSS). W. vol. Man.” Ne- gotiation J. K. engineering. and D. and D. Hipel (M’90–SM’92–F’96) received the Res. He has [15] . Syst. he was a Lecturer with the 159–177. pp. F. Li received the B.. design engineering from the University of Waterloo. University. Syst. dis.. He is currently a Professor of mathematics [29] L. 1994.. Canada. New York: Wiley. “Fuzzy moves. [32] L. Stanley Vineberg Memorial Visiting Professor- 1995.” Group Decision and degree in systems design.” Econo. 19–52. pp. “The decision support U. Academy of Engineering (CAE). Xiamen University. and D.degree in en- [25] M. 117–152. Theory of Games and Economic Waterloo. China. “Confrontation analysis as a diagnostic tool. Kilgour. [12] D. “Bulk water export issues in Canada research interests are in the intersection of mathematics. cluding mathematics. 30. pp. Hipel. and and the Lake Gisborne conflict. Princeton. and K. 1972. 126. Xiamen. 2002 Int. Man and Cybernetics (SMC) Society.” in Proc. He has held academic and administrative positions [27] M. 2nd Int. Howard. pp. Professional Engineer and has carried out consulting activities with engineering 2 [18] K. Osborne. Canadian putat.. vol. Hipel is a Fellow of the Royal Society of Canada (RSC). vol. Waterloo.Sc. pp. degree in systems engineering [5] N. as well as many uncertain ordinal preferences. and L. 489–507.. degree in applied math- games. degree in systems [6] S. Hipel. W. and N. Press. He is a recipient of the Distinguished Teacher [22] R. He is a Univ. “Modeling misperceptions in gineering physics.D.: EOLSS Publishers. ronmental engineering and sustainable development. 187–206. 4. Hipel. and X. the M. Fang. Cybern. Hipel. Wang. and Adjunct Professor vol. Res. Kilgour. pp. Waterloo. Kilgour. “Hypergames: The development of an approach to mod- eling conflicts. from a systems design engineering perspective. of Def. Hipel. D. L.D.” Behav. 56–66. pp. and the American Water Resources Associa- [21] H. voting.” Appl. Hipel. D. tice-Hall. and utilities in many countries. “Fuzzy approaches Science). 1997. Centre for Military Strategic and Disarmament Studies.” AWRA. and the Ph. “Utility representation of an incomplete preference relation. He is the author or co-author of four books. K. From 1994 to 2000. 1471–1475. Feb. no.. H.. Keith W. Canada. 54. no. 36. vol.” Appl.

in 1982.D. Toronto. PART A and the International Journal of Business Process Integration and Management. multiple criteria decision making. Prof. 34. degrees in Systems Design Engineering from the University of Waterloo.Sc.A.520 IEEE TRANSACTIONS ON SYSTEMS. Fang is an Associate Editor of the IEEE TRANSACTIONS ON SYSTEMS. and a Member of the Institute for Operations Research and the Management Sciences (INFORMS).Eng. NO. Canada. JULY 2004 Liping Fang (A’92–M’99–SM’00) received the B. VOL. particularly in interactive decision making. He has actively carried out research and consulting activities in the areas of industrial engineering. Tianjin. Waterloo. and decision making. and 1989. Canada. He is currently a Professor and Associate Chair of the Department of Mechanical and Industrial Engi- neering. 1985. China. degree in electrical engineering from Tianjin University. a Senior Member of the Institute of Indus- trial Engineers (IIE). AND CYBERNETICS. AND CYBERNETICS—PART A: SYSTEMS AND HUMANS. He is a registered Professional Engineer in the Province of Ontario. 4. University of Waterloo. . and Ph. He co-au- thored two books and is the co-editor of a book. and decision support systems. and Adjunct Professor in the Department of Systems De- sign Engineering. systems engineering. MAN. and the M. Ryerson University. Canada. engi- neering management. respectively. MAN.