EVOLUTIONARY SYSTEMS DESIGN
Tung X. Bui
The U.S. Naval Postgraduate School
Monterey, California 93943
Melvin F. Shakun
Leonard Stern School of Business
New York University
44 West 4th Street
New York, NY 10012-1126, USA
This paper presents a computer application for supporting negotiation. A negotiation accord is often the result of an intense, laborious and evolutionary negotiation process. During this process, disputing parties are confronted with goal, judgment and outcome conflict. This paper demonstrates the utility of a conflict resolution framework — Evolutionary Systems Design (ESD) — by using a Negotiation Support System. ESD seeks to guide negotiators to move their individual goals, judgments in such a way to enhance the chance of achieving a common solution . As illustrated by the use of NEGOTIATOR, a multi-attribute utility negotiation support system, we argue that computer mediation can prove to be an effective means to implement the ESD framework.
Keywords: Negotiation processes, Evolutionary Systems Design, NEGOTIATOR software
Negotiation may be characterized as "a process of potentially opportunistic interaction by which two or more parties, with some apparent conflict, seek to do better through jointly decided action than they could otherwise" (Lax and Sebenius, 1986). Negotiation involves both cooperation and conflict — cooperation to create value (increase the size of the pie) and conflict to claim it (take as big a slice of the pie as possible).
Research in negotiation has often focused on the quantitative and/or qualitative mechanics of seeking a consensus or a compromise outcome that would be acceptable to all. The traditional modeling approach treats a negotiation problem as a well structured problem and assumes a dynamic search for consensus within the context of the problem until a solution can be found. Such an approach presumes, however, that negotiation is a well-confined process. In fact, we argue that, more often than not, negotiation accord is the result of multiple, intense, and evolutionary negotiation processes, punctuated by streams of working agreements and disagreements. For each of these processes, a variety of process-dependent problem-solving methods can be used.
Recently, the much publicized GATT (General Agreement of Tariffs and Trade) accord in Geneva, Switzerland has been acclaimed by many high ranking government officials as "a milestone in the history of world trade" (Wall Street Journal, December 15, 1993; March 26, 1994). The accord was signed to prevent seven years of talking, threatening and cajoling from ending in collapse. From an analytical point of view, various compromises, breakthroughs and dodges were achieved through detailed analyses of the economic and social impacts of alternate solutions (e.g., revision of agricultural subsidies by the European Community to France in order to maintain the terms of agreement from a previous agricultural accord with the USA). From the point of view of problem representation and design, unsolvable issues such as financial services, free trade of "cultural goods" were left unsettled. Furthermore, resolved issues such as increased protection of patents are widely publicized to build momentum for the next rounds of negotiation which will likely include tougher issues.
The GATT case exemplifies a major aspect of negotiation processes, that is the dynamic and evolutive framework in which stakeholders identify negotiating issues, define preferences and make agreements. More important, it best illustrates the discontinuous or stop-and-go nature of negotiation. In this paper, we propose a methodological framework of negotiation that captures this negotiation reality, and illustrates that computer-based negotiation tools provides an evolutionary, integrated and continuous support to negotiation.
2. Negotiation: Values, Goals, Preferences and Problem Representation
We contend that a negotiation problem is a set of complex, self-organizing processes involving muticriteria, ill-structured, evolving, dynamic problems in which players both cooperate and conflict. These processes involve problem definition and solution.
Following our Evolutionary Systems Design (ESD) framework (Shakun, 1988, 1991), a problem involves the following sets of elements:
(1) values or broadly stated desires;
(2) operational goals or concrete expressions of these values;
(3) decisions, actions, or controls taken to achieve these goals;
(4) criteria based on goals for evaluating the effectiveness of decisions;
(5) individual preferences defined on criteria; and
(6) group or coalition preference defined on individual preferences.
More specifically, values are beliefs regarding desired modes of conduct and end-states of existence (Rokeach, 1973). For example, Maslow's (1954) values hierarchy involving safety, security, love, self-esteem, and self-actualization expresses and end-states of existence and terminal values. Furthermore, values and goals represent wants. In a typical negotiation, parties have to deal with both issues related to the way negotiation is handled (how), and to the exploration of consensual or compromise outcomes (what). During this search for a mutually satisfactorily solution, parties involved in a negotiation exhibit both rational and socio-emotional behaviors. The mixture of behaviors varies in intensity and composition as the parties go through different negotiation stages (Thomas, 1992).
Operational goals are beliefs defined by specific, unambiguous operations and are characterized by performance measures. They are operational expressions of higher level values. Goals are delivered by controls chosen by players. Goals are used as criteria for evaluating the effectiveness of decisions. When goals are perceived to be risky, criteria such as arithmetic means and standard deviations, can be defined on probabilistic goal outputs. Individual preferences and group or coalition preference are drivers in finding solutions.
We can therefore define a problem representation as one that provides more insights into the evolving relations among the above six sets of elements. We propose two hierarchies of relations (Shakun, 1991). As illustrated in Figure 1, hierarchy 1 relation is a framework for dynamically searching for the definition of the general problem. From a policy-making standpoint, values are expressed in the form of operational goals and realized by exercising control variables. On the other hand, hierarchy 2 relation is a framework for finding a solution. Together, hierarchies 1 and 2 define and solve an evolved problem in which policy making can be viewed as the process of delivering values to participants in the form of operational goals. Solution search involves problem restructuring and finding compromise solutions. A solution has been found when, in control, goal, criteria and individual and group preference spaces, the intersection of the coalition target (what the group wants) and coalition feasible technology (what it can do or get) is a single set or point.
Figure 1. Hierarchy 1: relation between control variables, goal variables and values
Coalition preference structure
(game theory, social choice, concession-making)
Individual preference structure
Figure 2. Hierarchy 2: relation between controls, goals, criteria, individual preferences, and coalition preference
3. An Evolutionary Framework for Negotiation Support
Typically, the negotiation process begins with a party's awareness of the conflict, either at the goal, judgment or solution levels.
Goal conflicts occur when a party seeks divergent or apparently incompatible outcomes, if necessary, at the expense of the other parties. Entering a bargaining session with the perception of incompatible goals often introduces a distributive bias. This bias has been shown to generate more hostility and mistrust between parties and diminish the number of suitable solutions generated (Bazerman, 1983).
Judgment conflicts differ from goal conflicts in that, while parties may share the same goal, they disagree over the best way of achieving it. Differences often reside in different interpretations of the same factual information (Bui, 1987). Parties may believe that they have information the others do not have. They presume that others may have come to an incorrect assessment and conclusion regarding that information (Thomas, 1992). Alternatively, they contend that, even if others possess adequate information, they simply use improper reasoning, do not understand the "true" problem or issues at stake, and tend to make wrong decisions.
Normative conflicts are manifested in a party's assessment and expectations on how the other party should behave (Keeney and Raiffa, 1991). Problems develop when one party is apparently perceived as violators of the standards or norms adopted by others. Negotiating parties who feel wronged by the violating party could experience feelings of disapproval, blame, anger, and hostility. This can escalate into sanctions to enforce conformity or to punish the other party, easily resulting in sub-optimal agreements or deadlock (Thomas and Pondy, 1977).
The acceptance of a solution is function of the extent to which protagonists' perceive that the proposed solution is a fair one. Conversely, the disagreement with a solution is function of the extent to which protagonists' perceive that the proposed solution is an unfair one. Again, fairness depends on the protagonists' rational-instrumental considerations and normative judgments. Normative criteria applied to conflict management involve the feelings of fairness and justice both as applied to distributive justice (the fairness of the ultimate settlement) and to procedural justice (the fairness of the procedure for arriving at the settlement) (Thomas, 1992). Perceptions of distributive justice are made up of several criteria, such as: (i) equity, (ii) consistency of results with similar conflicts, (iii) the relative needs of the parties, and (iv) consistency with accepted rules and norms. The perception that these criteria are satisfactory or fit within the party's allowable norms leads a party to view the outcome as acceptable. Normative procedural justice as identified by Sheppard (1984) and Thomas (1992) involves: (i) the neutrality of the third party, (ii) the ability of the principle parties to control the process, and (iii) protection of the rights of the principal parties. A party's perception of how they and the other parties are being treated during negotiations shapes reaction during the episode as well as affects party acceptance of a potential settlement. Thus, there could be some degree of ignorance in evaluating the other party's perception partly due to poor or no communication between parties.
In sum, negotiation dynamics could be characterized by the evolving relations, (goal/values and controls/goal relations in Figure 1), and by the five evolving spaces, (control, goal, criteria, individual preference, and group preference in Figure 2) in which antagonists try or refuse to resolve conflicts. This process is refined and validated through an evolutionary procedure, and is repeated until parties accept the outcome or break off negotiations. Supporting negotiations implies providing means that could help antagonists find the rightness in problem representation, and negotiation solution (Hierarchies 1 and 2).
4. Supporting Evolutive Negotiation with Computer Mediation
Negotiation Support Systems (NSS) are computer assistance for negotiations. Research on NSS has primarily focused on two key technological aspects: (i) group decision and/or conflict resolution models to help negotiators reduce discord and increase the chance of reaching consensus, and (ii) providing rich communications media to enhance communication exchange between antagonists. Computerized models of negotiation originate from a number of disciplines. Operations research and management science, economics and applied artificial intelligence are major disciplines that contribute to the formulation of negotiation models and processes. Models found in the NSS-related literature include those derived from game theory, multiple objective optimization, and rule-based advisory systems (Bui, 1987, 1993).
Computer support can be used to assist the negotiators in interactive information elicitation and process them in an orderly manner.It is not unusual that negotiating parties define the wrong problem. Shakun (1992) suggests that rightness in problem representation requires rightness in the relations in hierarchies 1 and 2 constituting that representation. To support finding the right problem, an NSS can help negotiators view the problem in a transparent and structured manner. Transparency refers to the ability of the decision maker to define, understand and assess the problem. Via user-friendly interface and structured modeling and representation, the decision maker has a better chance to clarify not only his/her problem but the one of his antagonists as well. Structuredness refers to the extent to which the problem is formulated in a systematic manner (e.g., using tables, graphical representations, formulas). By imposing a certain level of structuredness in problem formulation, NSS can also be used as a shared and common language for mutual understanding (Rhee et al., 1995). Structure can help the negotiators appreciate better the strengths and weaknesses of the other party's position and arguments (for example, see Thomas, 1992). A joint and open modeling effort may be to the advantage of all parties.
Along with the ESD framework, NSS can be used to provide continuity to negotiation. Observations of real life negotiations that have ended with successful outcome, demonstrate that temporary interruptions of the negotiation process can help enhance the chance of reaching consensus (Lax and Sebenius, 1986). A break can be useful in forcing highly emotional parties to calm down, thus enhancing the feel for problem rightness. However, discontinuities of negotiation, especially those that are rather long, can be detrimental to negotiation outcomes. Motivation becomes diminished, problems and issues are forgotten and the urgency of findings a solution gone. In an NSS environment, with enterprise-wide networking, information exchange can be supported via communications channels thus reducing the time and geographic barriers that separate negotiating parties (Binbagliosu et al., 1995; Bui et al., 1995). Also, with structured modeling methods implemented in fast computing working environment, more transparent screening of proposals and working agreements can be achieved, allowing negotiators to "navigate" quickly in hierarchies 1 and 2. This process not only helps negotiators increase the chance of finding a compromise solution, but frequently guides them in reaching a better-than-expected solution.
5. NEGOTIATOR:An Interactive Procedure for Negotiation
The evolutive approach to designing negotiation support systems can be illustrated by the implementation of NEGOTIATOR, an NSS installed in a network of personal computers using MS-Windows for Workgroup and equipped with multi-media and video-conferencing capabilities. Using a multi-attribute utility model, NEGOTIATOR allows negotiating parties to evolve through hierarchies 1 and 2. Each party can have its own computer support environment that contains models customized to its needs. The environment describes the issues in which NEGOTIATOR allows the negotiators to engage in a joint and open modeling effort.
In practice, technical experts and advisors supply the bulk of the information to the negotiators either before or during the negotiation process. Even if such information is accurate and complete, there is no reason why the negotiators themselves could not exercise their freedom of choice at the time of negotiation through joint concession, and experimentation with simpler models of their own. NEGOTIATOR allows negotiating parties to navigate dynamically through the relations in hierarchy 1 and the five dimensions in hierarchy 2 in search of a solution. The Evolutionary System Design framework is realized by helping negotiators focus on asymmetries of interests between the parties so that the terms of the final treaty are better for both (Barclay and Peterson, 1976). A good treaty is one that yields to each party those issues which are more important to it. Thus the two parties should try to push the negotiation toward the Pareto optimum by capitalizing on asymmetries of interest, and whenever possible by redefining the situation to reveal more asymmetries. A treaty is Pareto optimum when it is not possible to increase the utility of one party without decreasing utility of the other (Bui, 1990).
The essence of the procedure is described below.
Step 1. Identify values, goal variables in hierarchy 1 associated with the major agreements that the parties seek to sign.
Step 2. For each of the agreements being considered, identify a common set of major issues (control variables in hierarchy 1) about which the parties may disagree.
Step 3. Each party assigns relative weights to each of the issues (individual preference structure in hierarchy 2).
Step 4. Define the range of values for all the issues as identified by both parties (levels of control variables in hierarchy 2). As the parties enter the negotiation, they offer their initial positions with regard to each of the issues enumerated.
Step 5. For each party, determine individual-issue weighted utility curves (individual preference structure in hierarchy 2). The determination is made by taking the product of the utility values and the respective relative weights of the issues.
Step 6. For each issue, compute joint utilities by aggregating the weighted utility functions of the parties (coalition preference in hierarchy 2). The aggregation could theoretically take any mathematical form. The simplest form is additive. For each of the issue, choose the term that corresponds to the highest point of the joint utility curve.
Step 7. Based on the terms of the issues suggested in 6, determine the total utility for each party across all the issues.
Step 8. Search for improvements and restructuring. The concept of joint utility allows for the possibility to check for non-cooperative issues and suggests restructuring. A cooperative situation is one in which the highest value of the joint utility curve is higher than the individual maximum utility values of both parties. Conversely, a non-cooperative situation is the one in which the highest value of the joint utility curve corresponds to the highest for only one of the parties, leading to unbalanced treaties. In this circumstance, it is recommended that the single non-cooperative issues be split (restructured) into subsets of more cooperative (asymmetrical) issues.
As illustrated by the example in Section 6, NEGOTIATOR is designed to support the improvement and restructuring process. It provides the user with simultaneous displays and print-outs of utility graphs, negotiation results in tabular forms, and with a spreadsheet to perform sensitivity analysis on the data suggested by the NSS or the modifications requested by different parties. Under a multi-tasking environment, multiple sessions of NEGOTIATOR can be run, allowing users to conduct parallel bargaining.
6. An Example
Figure 3 illustrates the ESD concept with NEGOTIATOR. The example is adapted from a real-life labor negotiation. The labor union of a middle-size factory which produces electronics components is seeking new terms and conditions of the labor contract with their management counterpart. As a result of multiple meetings between the labor union committee and its members, the labor union (party A) initiated a request to the company management. Three salient aspects were identified: salary increase (5% increase), duration of labor contract (maintain the existing 2-year length) and duration of vacation (maintain the 4-week condition) (Figure 3a).
The problem triggered by the labor union has forced Management (party B) to take a position. They studied the three issues addressed by the union and informed the latter that they are willing to engage in negotiation if the union is willing to consider productivity as part of the negotiation. In fact, management has recently discovered that increasing the quality of their products while reducing some production costs would be their only approach to survive a fierce competition in a global market (Figure 3b). The four issues form then the first collective goal space. As such, the goal space in NEGOTIATOR can be viewed as an aggregation of the spaces of the two negotiation parties. Note that in this negotiation, the goal space is also the control (decision) space.
The management proposes a freeze in pay, a 6-month labor contract with 3-week annual vacation, and requests that productivity must be increased by at least 8%. Reacting to the proposal, the union revised its starting position(Figure 3c). Based on these starting positions, the two parties begin to analyze the problem. Figure 3d to 3g respectively show the parties’ weights on issues, and a sample of their utility curves for the issue of salary raise.
Three solutions are proposed: (i) highest joint utility, (ii) mid-point solution, and (iii) relative importance (Figure 3h). The first proposed solution yields the highest possible joint utility, i.e., 136 points total for both parties. Another solution is based on the mid-point principle that yields a joint utility of 97. As its name suggests, the mid-point principle is one that finds the solutions by equally splitting the terms requested by the negotiators. For example, the mid-point principle suggests that the term for duration of the contract is 15 months (Figure 3h). Fifteen months is the mid-point of the management’s 6-month proposition (Figure 3b) and the union’s 24-month proposition (Figure 3c). The third solution is based on the concept of relative importance that gives each party what he wants on those issues for which his relative importance is larger than that of the other party. The relative importance concept suggests a solution whose terms yield a joint utility of 132.
The solutions suggested by NEGOTIATOR in Figure 3h are however only a basis for evolutive exploration of new, and hopefully, better solutions. The numbers of issues, issue weights and utility values can be refined or modified till new and more satisfactory solutions can be found.
In fact, the highest joint utility solution proposed by the NSS at the first round of negotiation was not well perceived by the union. Although the solution proposes a 4-week vacation that is what it wanted, no salary raise is recommended. Furthermore, the management seems to come out winner for a total utility of 75 versus 61. This discontent is further substantiated by a close examination of the issue utilities. While both parties seem to have found a compromise on productivity (cooperative issue as shown in Figure 3j with a joint utility curve of convex shape), the salary issue (Figure 3i) clearly went in favor of the management. In return, the labor union obtained almost what it wanted for the duration of the contract.
Having expressed their disagreement, union leaders enter the second round of negotiation by arguing for a 24-month contract and a recognition of their effort to increase productivity (5%) by a "modest" salary increase (4%). The scratchpad of the NSS (Figure 3k) is used to enter the new issue values and derive corresponding utilities. The results reflect the compromise solution with a lower total utility (128), but with a less uneven utility ratio between the two parties.
Management reminds the union that they had a good compromise on productivity (6%), along with the vacation issue (4 weeks, Figure 3k), and suggests that these issues not be re-considered. On the other hand, it is willing to restructure the problem by linking the union’s desire to have longer labor contract with a two-phase salary adjustment. The ESD heuristic referral process could be used to support this restructuring (see Shakun, 1995). A new problem representation is suggested in the third round in which parties are required the parties to (re-)assess their preferences (e.g., utilities functions shown in Figure 3m and 3n). A new highest joint solution is found (Figure 3l): 24-month contract, with a 1% pay raise with immediate effect, and another 3% a year later. This solution seems equitable (68/66). Together with the two issues agreed earlier (i.e., productivity at 6% and 4-week vacation in Figure 3k), the total utilities are 95 and 93 for the union and the management respectively.
7. Concluding Comments
The ESD framework, as illustrated by the labor dispute case, contends that the evolving group or joint problem representation is based on confronting individual problem representations. If the latter are not fully shared by individuals in the group, the public group problem representation will be incomplete. Each party privately can subjectively estimate missing information — i.e., establish his/her private group problem representation. In NEGOTIATOR, this estimate can be achieved by simulating the behavior of the other party.
ESD involves evolution of the group problem representation, that is, evolution of its relations. It promotes (i) problem adaptation through information-sharing and concession-making; and (ii) problem restructuring or reframing. The problem adaptation and structuring can be modeled by mathematical relations expressing hierarchies 1 and 2. In this sense, when satisfactory solutions are not forthcoming, problem restructuring is a key approach.
In an organizational context, negotiation is often an integral part of policy making. In this paper, policy making is viewed as a series of processes that involve multi-criteria, ill-defined, evolving dynamic problems in which decision makers both cooperate and conflict. Thus, for an organization, negotiation is a continuous strategic effort to deal with all immediate and potential partners.
Supporting continuous navigation through the evolving problem representation (hierarchies 1 and 2) using a computer-based negotiation support system is expected involved parties to explore quickly new collaboration opportunities. NEGOTIATOR, by
(i) establishing a consensual database as a foundation for negotiation,
(ii) evaluating the impact of perceived uncertainty,
(iii) providing communication links for bargaining and discussion,
(iv) suggesting restructuration of non-cooperative issues, and
(v) helping search for agreements moving towards Pareto-optimality,
provides a model-based interactive facilitation process. This process offers a comprehensive framework to allow the parties to concentrate on joint problem-solving rather than on arguing verbally.
Barclay, S. and Peterson, C.R., 1976, "Multi-attribute Utility Models for Negotiation," Decisions and Design, Inc,. McLean, Virginia.
Bazerman, S., 1983, "Negotiator’s Judgment: A Critical Look at the Rationality Assumption," American Behavioral Scientist, Vol. 27.
Binbasioglu, M, Bui T. and Ma P.C., "An Action-Resrouce Language for Argumentation: The Case of Softwood Lumber Negotiation", Proceedings of the 28-th Hawaii International Conference on Systems Sciences, IEEE Society Press, Vol. 4, pp. 262-269.
Bui, T. and T. Sivasankaran, "Fuzzy Preferences in Bilateral Negotiation Support Systems," Proceedings of the 24th Hawaii International Conference on Systems Sciences, IEEE Society Press, Vol. 4., pp. 687-694.
Bui, T., 1993, Designing Multiple Criteria Negotiations Support Systems: Frameworks, Issues and Implementation," in Tzeng et al., MCDM: Expand and Enrich the Domains of Thinking and Applications, Lecture Notes in Mathematics and Economical Sciences, Springer Verlag.
Bui, T., Ma P.C. and Stricker C., "Supporting Argumentation in Software Development", working paper, Department of Information and Systems Management, Hong Kong University of Science and Technology, 1995.
Keeney, R.L., and Raiffa H., 1991, "Structuring and Analyzing Values for Multiple-Issue Negotiations", in H.P. Young (ed.), Negotiation Analysis, Ann Arbor, I, University of Michigan Press, pp. 131-151.
Lax D.A. and Sebenius J.K. (1986), The Manager as Negotiator: Bargaining for Cooperation and Competitive Gain, Free Press.
Maslow, A.G., 1954, Motivation and Personality, Harper and Row, New York,
Rhee, H-S., Pirkul H., Jacob V. and Barhki R., "Effects of Computer-Mediated Communication on Group Negotiation: An Empirical Study", Proceedings of the 28th Hawaii International Conference on Systems Sciences, IEEE Society Press, Vol. 4., pp. 270-279.
Rokeach, M, 1973, The Nature of Human Values, Free Press, New York.
Shakun, M.F., 1988, Evolutionary Systems Design: Policy Making under Complexity and Group Decision Support Systems, Holden-Day, Oakland, California.
Shakun, M.F., 1990, "Group Decision and Negotiation Support in Evolving Nonshared Information Contexts," Theory and Decision, 28 (3), 275-288.
Shakun, M.F., 1991, "Airline Buyout: Evolutionary Systems Design and Problem Restructuring in Group Decision and Negotiation," Management Science, 37(10), 1291-1303.
Shakun, M.F., 1992, "Defining a Right Problem in Group Decision and Negotiation: Feeling and Evolutionary Generating Procedures," Group Decision and Negotiation, 1(1), 27-40.
Shakun, M.F., 1995, "Evolutionary Systems Design: A Formal Modeling Framework for Task-Oriented Group Processes," Group Decision and Negotiation, forthcoming.
Sheppard, B.H., 1984, "Third Party Conflict Intervention: A Procedural Framework," in Staw and Cummings (Eds.), Research in Organizational Behavior, Vol. 6, JAI Press, Greenwich, CT.
Thomas, K, 1992, "Conflict and Negotiation Processes in Organizations" in Dunnette (Ed.), Handbook of Industrial and Organizational Psychology, Consulting Psychologists Press, Palo Alto, CA.
Thomas, K. and Pondy, 1977, "Toward an ‘Intent’ Model of Conflict Management Among Principle Parties," Human Relations, Vol. 30.
APPENDIX A. Overview of NEGOTIATOR: A Multi-Attribute Negotiation Support System
NEGOTIATOR is a NSS for MS-Windows. Installed in a network of MS/DOS-Windows personal computers, the software is a Windows-based NSS that can accommodate several negotiating parties in a face-to-face (i.e., decision room setting) or a distributed conflict resolution process.
Table 1. Evolutionary System Design and Conflict Resolution