From: AAAI Technical Report WS-94-04. Compilation copyright © 1994, AAAI (www.aaai.org). All rights reserved. 122 The GenericMetamodel,the Conflict ModellingCycle and Decision Support M.I.Yolles Liverpool John MooresUniversity Business School 19 April 1994 Abstract Metamodelsprovide a mechanismfor guidance in modelling. They offer a structured approach,which is appropriatefor the modellingof situations and processes. An exampleof a metamodel is a modellingcycle, and one is proposedsuitable for conflict processesin groups with critical size. The wayin whichsuch a modellingcycle can be implementedon a computersystemfor decision support is discussed. 1. Introduction This paper is concernedwith the modellingprocess of conflict which arises from change, and the simulation of that process. The modelling of modellingprocesses is a metamodellingprocess, providing guidanceand structured reasoning to situations that maybe somewhat messyin the way in whichthey are defined. Thus, this paper is concernedwith metamodelling, andin particular it concernsthe structuredmodelling of conflict arising fromchangefor large groups.It also has interest in the computerisationof such modelling. Whenattempting to model, and thus examining the conflict processesof groups,it is appropriateto differentiate betweenthose whichare small and those whichare large: smallgroupstend tO, be illstructured in their communications and ffecision processes, that is they do not conform to a predeterminable pattern or relate groupentities in a predeterminable way. Larger groups tend to be morestructured, havingformalised processes and entity relationships that are better known and more predeterminable. Thus, the structural nature of groupsunderchangeitself changesaccordingto the size of the group within whichit occurs. Small group processes tend to operate differently from those of large groups. They mostly operate informally and are unslructured. A group can be thought of as becominga large group whenit has acquired a critical massof people. Like so many examplesin real life that can be described as having an instantaneous metamorphosis[Thom, 1975],the critical changethat distinguishessmall group processes from large group ones maywell appear to be sudden and distinct. However,as groupsget larger, so groupnormsstart to appear;as they increasein size, so too doesthe complexityof their relationships, communications,and other processes. Withthis formalisedprocessesstart to develop, and the group thus becomes more structured andmoreeasily representableby formal models. Models can be classified on a hard to soft continuum.In hard modelsthings tend to dominate a problem andits setting, whilein soft modelsit is peopleandtheir psychologicalneedsthat dominate. In very soft contexts, the modelmaybecomethe activity. For example the named conceptual domains of Systems Engineering, Project Management,SystemsAnalysis~ and Operational Research are closer to the hard end of the continuumwhile ManagementCybernetics, Soft Systems Methodology, and Organisational Developmentare closer to the soft end. Harder approachestend to adoptmoreexternally structured elementswithin their operational frameworks than the softer approaches which tend to be more unstructured. Domains that are highly structured haveelementsthat are explicitly well defined and can more easily be modelled. Thus, in systems engineering it is the norm to modeland where possible test a solution before implementingit, while in OrganisationalDevelopment, the only way to test a modelis to experienceit. Theapproachto modellingconflict processesis frequently better undertaken whenguidance in creating and validating modelsis provided. This representsa structuredapproach.In particular, it is important that the approach to a problembeing modelledis to be well structured. This can often be accomplished throughthe use of a modellingcycle, and modelling cycles are representative of metamodels. One purpose of this paper is to consider a modellingcycle whichis directed at large groupconflicts. Whenconsidering computeraided application of a metamodelto a problemdomain,the softer the problem, the more difficult it is to generate a computer system able to adequately deal with general modelling and simulation requirements becauseof the needfor machineintelligence, level of knowledge, and decision makingfacility needed. 123 2. ModellingCycles Simon [1960] was a major contributor to ManagementScience. His concern lay in the development of a decision science, and in order to do this he expressed the prevalent ideas on modelling developmentas a general cycle for decision makingprocesses under goal seeking behaviour.Its three phasesare Intelligence,Design, and Choice which can be iterated through to progress a problem.The cycle provides guidance in modellingdecision problemswhichtells us that problem domains must be properly examined, options identified, and models generated and applied to the domains. The Simoncycle has provided guidanceto the process of decision making.However,it does not provide modelbuilders with direction about model building techniques or approaches,nor provide a philosophical orientation for so doing. A development of this cycle was suggested by Rubensteinand Haberstroh[1965], and a variation designed for computer software developers was later produced[Sprague, 1986]in order to tackle well structured problemdomains.Thelatter hard approach come out of the stable of Systems Development Life Cycle, and this type of approach has been used in situations in whichthe analyst does not envisagethe involvementof people while addressing the problem domain. It therefore providesfor a highlystructuredmodellingai~proach that can be represented in terms of well defined harder modelling techniques. To tackle more unstructuredprocessesinvolvingpeoplea different needarises. One well knownmetamodel[Checkland, 1981; Checkland,Scholes, 1990] called Soft Systems Methodology (SSM) has been used to solve unstructuredsmall groupdynamicchangeproblems. Thephilosophicalstand of SSM,however,is strong is demandingconformityto the wayin whichthe cycle operates, and suggests that consistent with soft perceptions, solutions to problemscannot be modelledbut rather mustbe experienced. Someauthors perceive that while SSMprovides a soundapproachtowardsthe solution of problems of change,it is constrainedin its breadth,since its philosophyof operationis very specific. Floodand Jackson[1990] have defined their moreflexible metamodel which introduces the idea of a "metaphor" to representan analogousconceptto the problem in hand. The metaphor thus helps to identify the context of a situation. This approach has beencalled Total SystemsIntervention (TSI), andconsists of a cycle of three phases:Creativity: use systemmetaphorsas organisingstructures; e.g. see an organisation as a machine(closed system), an organism (open system), a brain (learning system), culture (norms, values), team (unitary political system), coalition (pluralist political system), or prison (coercive political system). Outcomeis the dominantmetaphor.Choice:select an interventionstrategy or set of methodologies as appropriate.Useanyof the tools available fromthe hard-soft continuumof techniques. A dominant technology may be found. Implementation: employs a particular system methodology to translate the dominant visionof the organisation,its structure, and the general orientation adoptedto concernsand problemsinto specific proposals for change. Common to the above metamodels,and indeed others, it is possibleto generatea definitionfor a generic modellingcycle. Thethree phasesthat are defined are Analysis, Synthesis, and Choice. Analysisis the breakingdownof a probleminto its components, includingits context,the identification of its structures, andits orientations.Synthesisis the building up of a set of componentsinto a coherent picture, from the integration of ideas derivedfromthe analysis, to the constructionof the prerequisites for a model. Finally, Choice is anythingthat involves the selection of something, including implementation. It is quite a simplematter to apply the generic modelto any modellingcycle from the very hard SystemsDevelopment Life Cycle, to the quite soft Organisational Development cycle. The fundamental distinction between the different metamodels then becomes the philosophical approach.In principle, one can consider that the generic metamodel is held on a slide on the hardsoft continuum on which reside a variety of modelling/problemstructuring tools, and as it movesit picks up those modellingtools it requires and appropriate philosophical approaches. Having completed a cycle, it is possible to moveto another division on the continuum,also perhapsadoptinga newphilosophical position. This approach must implicitlyholdflexibility in its philosophical stance. 3. The Conflict ModellingCycle In the social sciences,it is frequentlythe case that modelsare built in a waywhichis not structured. Theycouldthereforebenefit fromthe applicationof a metamodel.Modelsfor such processes mayvary in their position on the hard-soft continuum. Whilethe TSI approach could be useful here since it providessomelevel of flexibility, it does not appearto give guidancein providinga wayof directly connectingthe examinationof the problem 124 Context analysis ~ f Structur, d {/~-~ system _* N~-~fluence a~ecstI°rY Outcome evaluation ~ analysis ~ C~C-rf ~ ~R/gffON / ~ ’ Mode ©~oic~ se~)eclnon I Smbillty analysis M delN~~Acflv pmro "~c~a’~l~ng Structured synthesis Propositional synthesis ~ Pruning Alternatives synthesis Fig.1 Conflict Modelling Synthesis Cycle domain with traditional hard madelling approachesas might be required in the semistructured problemsof large group conflicts. In an attempt to address this type of problem domain, the conflict modelling cycle (CMC) presented in figure 1 adopts the three phases of the generic model. It is thus consistent with the Simon cycle, and broadly SSM and TSI, though necessarily the terminology and philosophy have changed. In CMC,the first phase Analysis involves domain examination and evaluation. The second phase is Structured Synthesis, and involves propositional definition of the unstructured problemdomain, and the structuring of decision alternatives. Modelling Choicesis the third phase; it is concernedwith the selection/implementation of modelling approaches which may be soft or hard. In the case of soft modelling the model defines the approach to a solution. In the case of hard modelsthen this phase in addition addresses explicit modelselection and evaluation, and ensures consistency between the propositional base of each of a possible set of modelsavailable for selection, and the propositions defined in phase 2. 3.1. Systems Analysis Systems analysis requires that the problem domain is defined, examined, and analysed. Analysis is concerned with examining the system situation for actors and conditions which relate to the resultant states that mustbe defined and selected; data inputs are obtained, processed, and examinedfor clues that mayidentify problems or concemedwith gathering data, identifying objectives, diagnosing problems, validating data, and structuring problems and environments. Systems may also be dynamically and structurally stable. Dynamicstability relates to the movementof a system in its achievement of desired goals. Here, a desired goal will be achieved as time passes if the systemis dynamicallystable. Necessarily, therefore, dynamicstability is time related. If a system has reached a goal and adheres to it, then the system has achieved a condition of equilibrium. Structural stability is more properly referred to as structural criticality. A systemwhich is close to criticality mayreact significantly to a small change in one of its parameters. Under change, a system may pass through a critical condition, whenits structural relationships alter. A big change caused by a small event shows the 125 systemnot to be stable. In terms of the conflict modellingcy¢’le, the phaseelementsare defined in the followingway’P1.1 Context analysis Examinethe nature and context of the conflict in general terms, and the environment in whichit operates. Identify the type of problem, i.e. hard (quantifiable or single solution) or soft (unclear solution), bounded unbounded.Use techniques like mindmaps, spay diagrams, force field analysis, relationship diagrams. P1.2 Structural analysis Define the system at various levels; use system maptechniques and define the nature andboundariesof actors andother entities. Usesystemsmethodology to identifying domainstructure andits description(e.g. boundary parameters, degree of boundaryfuzziness, system entropy); identify actor problemperspectives. Identify generalgoals of the actor systems. P1.3 Trajectory analysis Trajectory analysis is concerned with problem definition and the movements that are madein solving it. Problem definition: Theproblemdomainis the problemand its set of actors, parameters, variables, and constraints.Cleardefinitionof this canbe difficult whenthere is sufficient complexity.In reducing complexity one then might: (a) examine the changes that mayhave invoked the problem, Co) identify the problemboundaries and associated parameters, (c) examinehowmanyproblemsexist and whetherthey can be reducedto discrete subproblems, (d) examinepossible problemsolving schedules.Trajectorydefinition: Eachparticipant in the conflict is an actor with a frameworkof perception, perspectives to the problem, and decisions and actions taken which constitute a pathway through the domain. The pathwaywill havea direction which,if intended, representsthe aimof the process,andidentifies a set of vectorsof movement.The trajectory is the set of vectors takenwiththe resultant goals that maybe achieved. The difference betweenan intended and an actual trajectoryis an indicatorof howstable the situation is. Thus, if measuresof achievableand intended goals, aimsand objectives can be made,a measure of stability can be found. Therationale andfeasibility of actor trajectories should be evaluated in relation to earlier phase elements. Techniqueslike objective trees, and multiple cause diagramscan be used. P1.4 Influence analysis Establish relationship between entities within the system and its environment. Use techniques like influence diagrams. This phase should reflect on P1.3 by suggestinginfluencesfor trajectory changes. Toundertakeanalysis, it is essential that actors and their influences are adequately understood. Actors havegoals, objectives, strategies, and an external environmentwith whichthey interact. Theyhaveinternal constraints as well as external ones, variables which include general cultural attributes. This applies to all classes of actor, whetherthey are enterprises, cultural groups, or nationstates. Culture, defined in terms of its attributes, determines the value attached to data and information[Yolles, 1992]. 3.2 Alternatives Synthesis By alternatives synthesis is meant selecting, inventing, or developing possible options or decisionscenarios.It is effectivelya designphase, and the development of decision scenario alternatives requires an adequateknowledge of the problemarea and an ability to generate feasible alternatives. In the modellingcycle, alternatives synthesis includes the followingelements:P2.1 Proposition synthesis Formally define propositionsrelating to the situation. This can be important because parties to a conflict operate within their owndistinct frameworks;whetherone party acts rationally maydependvery muchupon the nature of the framework, andif conflicts are to reachsatisfactory conclusions,it mustbe possible to mapfrom one frameworkto another coherently. This in essencedefines a principle of interactive relatively for the parties to the conflict. Propositional definition may usefully use established terminology (e.g. game theory) introducingterms like player rationality, closed games,stationary games.Alternatively in a softer approach,it can simplylist the set of assumptions that are madein order to construct a modelling approach. P2.2 SynthesisingAlternative Structures Generate a range of options: these include player participation, and possible player choices which maybe defined for the situation; include possible player state choicesthat maybe feasible andplayer coalitions in an n-player system. This involves modelling interactive player relationships as definitive scenario possibilities. Decisiontable techniques are appropriate, as might be other decision related approaches like Pugh’s matrix [1984] often associated with the Organisational Developmentmethodology. P2.3 Pruning Pruning is the reduction of the alternatives identified in P2.2. These will be unstablescenarioswhich,by their elimination,will reducethe alternatives to a core set of Optional 126 Reality State (ORS)scenarios. ORSare perceived options or modesof adaptation within the system, andrepresentthe possible states in a systemswhich contributeto the definitionof the systemstructure. Usetechniqueslike expert evaluation or Conflict Analysistheory. In general, this phase is concernedwith the manipulation of data, quantifying objectives, generatingreports, generatingalternative scenarios, assigninguncertaintiesor valuesto alternatives. In very soft applications, it relates to reducingthe tactical optionsavailable to a core approach. 3.3 Modelling Choices Modellingchoices involves identifying/selecting modelscapable of representing feasible decision scenariosfromthose optionsavailable. It includes the evaluationof modeloptionsandtheir ability to represent player environments and decision scenarios, and examinationof the consequencesof modelling option in respect of a changing environment. It is necessary to activate these modelsas a solution to the problem. This will generate outcomeswhichmaybe possible problem outcomes. The modelscan be validated and the outcomes evaluated in terms of the problem domain.In termsof the modellingcycle, this phase includes:P3.1 ModelSelection Identify modellingoptions and approaches; identify methodsfor modelling perceivedand hypothetical situations whichmight also model the future; provide the choice of selecting modelalternatives; identify the model demands,constraints andperspectives explicitly; define modelpropositions, stability mechanisms, and propositional base including any normative assumptions.Evaluate models,comparingselected modelsto the real worldsituation, and identifying convergencebetweenmodelling options and real worldsituation - that is convergence with P2.1. In very soft applications,this step simplyrelates to adoptingthe core approachof P2.3. P3.2 Activate Modelling Process Involve ORS scenarios from P2.3, and set up a modelling technique. Identify any additional parameter estimationor variable estimationrequirementsbased on historical evidence,assembledata, and prepare process modellingcomponents. Generate modelling results. In the case of stochastic modellingprocesses, approacheslike MonteCarlo simulation, Markovprocesses, or Weibullgames(see section 6) can be adopted, and perhaps compared. The application of models involving normativemechanisms must be connected with P2.1. In soft problemsthis step relates to starting the experience. P3.3 Stability Analysis Investigate dynamicand structural stability of the synthesisedsystem.Use mechanismsimplicit to modelswhere possible. Examineconvergencewith P1.3. P3.4 Outcome Evaluations Validate model: examinethe selected modeloutput and comparethis to actual events. In relatively hard situations this maywell involve a quantitative approach, and in the case of there being numericaloutputs these mustbe interpreted qualitatively. Asoft approach requires checking that the progress of the experienceis appropriate. A matchbetweenmodel outputs and acceptable or real world events will indicate the level of ability of the modelling approach. Thus, relate modeloutcomesto phase P1.3. If convergenceoccurs, identify impact of modellingprocessby scenario adaptationin phases PI.1, P1.2, P1.4. Otherwisereturn to P3.1. This phasedistinguishesthe ability of eachmodel to representthe situation andthe constraints under which it operates. Validation of a model only occursif the modellingoption evaluationhas been successful. 4. ModellingDecision Options The modelling cycle requires a full systemic evaluationof the conflict domainsatisfying each phase of the Conflict Modelling Cycle. In Analysis, the problemenvironment is very strictly viewedas a system whichis formally defined in termsof the appropriatetheory; thus the constituent elementsof the problemenvironmentare defined, includingwhothe participating players are, their attributes, functions,andrelationships. It defines the frameworkof the situation being studied for eachplayer: that is, identificationof the natureof a player and its boundariesand influences. This is equivalentto defining the framework of perception of players, and very closely relates to the propositions that determinethe wayin whichthey operate. Correctly, data should be collected from all players within the defined system and differentiation shouldbe madebetweenobservations and player perceptionsin order to addressthe next phase. In the secondphase, Synthesis, a formalset of propositions are synthesised whichrelate to the analysis of the conflictual system. For instance howfar can one assume player rationality and within whatcontexts. Thepropositional base may include, for instance, considerationof underwhat conditionsplayerswill act against their rationally establishedpreferences. 127 This can be followedby the creation of a set of decision tables. Eachdecisiontable will consist of three connectedsub-tables: .(a) properties, (b) objectives, and (c) goals [Yolles, 1992]. In the context of social conflicts, properties formthe currentcharacteristicsof a player, andrelate to its power base over each of the social, economic, political, andcultural domains.Goalsrepresentthat whicheach player intends to achieve in the long term. Objectives are the set of decision 9ptions available to the players whichnormallyrelate to their rationally establishedpreferences.Asa result of Decisiontable analysis, feasible solution can be found for the problemunderconsideration. Feasiblesolutions are those whichare logically consistent. Feasible ORSare those ORSstates whichcan logically exist, and are determinedfrom the set Of systemproperties. Theset of feasible ORSsin an n-player system composepossible scenariosof interaction for examinationwithin a decision makingframework.Thus, each scenario becomesa feasible decision makingsolution of the problemenvironment. A decision table would be created for each of 4 dimensionsof concernin the aboveexample,that is political, social, economic and cultural [Kemp,Yolles, 1992]. Scenarios are established within these tables. Ascenario maybe a oneplayer identity of ORSthat is feasible under definedconditions. Theseconditionsare definedin the first part of the table. A scenario mayalso representinteractionwhenat least twoplayers(e.g. an ethnic groups and a host player, or two coincidentcultural minoritygroups)defininga set of states in contrapositionto eachother acrosseach of the 4 domains.In this case the scenario also represents a set of mutualpositions taken by each playeron eachstate in the interaction. 5. Developinga Conflict Tableau In consultation with a colleague G.Kemp,it was found that the decision table approach can be developedfurther as a modellingtool in its own right within the conflict modellingcycle. Here, objectives fromeach player havebeenassembledin order to form a conflict tableau whichcan be formulated into a set of feasible interactive scenarios(see fig. 2 and3). WemaychooserSp to represent the rth decision table of player p, and’s~ to representthe combined objectivestables for all of the playersat the rth iteration at the mstable decision options. Wecan nowdefine a futures trajectory whichmayoccur alonganybranchof the futures tree. Thefuturescenarioset’sj for the rth iteration for the jth scenario is generatedby inspectionthrough the initial use of the methodology describedbelow. Initially the ~Spdecision tables for r=0 are generatedwithin the modellingcycle, anda tableau ’s i with j scenario possibilities is created. This tableau enables an interactive evaluation of the conflict domainto be attempted. Oncethis has beenpruned,an investigation of howselection can effect the 1Sp decision tables will be examined within the modellingcycle. It maybe that there is no difference between °Sp and ISp, whena new futures set ls~ is generateddirectly. Consider each objective table as a possible outcome of a decision option. A number of different outcomescan developwhichare presented within a conflict tableau. Eachof these possible outcomesis as a scenario. For an exampleof this see [Yolles, 1992a]. 6. Choosing a Model Thethird phase of the modellingcycle is that of Choice,wherethe modellingapproachis chosento solve the perceivedproblem.Twoof the features of the approachadoptedwithin this proposalare: 1) use the goals table for examining dynamic stability duringiteration withinthe modelling cycle; 2) use the objectivestable withina conflict tableau to examinethe structural stability of the system under examination. Theidentification of a stable set of ORSreduces the size of the conflict tableau. This reducedset can be usedto evaluatethe impactof each optional scenarioonthe original propertiestable to create a set of possible futures as shownin fig. 2. This in turn enablesa dynamicstability evaluationto occur on possiblefutures. The need to examinethe conflict tableau to investigate the structural stability of the conflict environment requires the use of a methodology, the propositional basis of whichconforms to that determined for the system overall. Examplesof some of the methodologies that might be appropriate are in particular Conflict Analysis [Fraser, Hipel, 1984], or a variation on Saaty’s multivariate decision analysis approach(called Analytic Hierarchy Process [Zahedi, 1986]), or simplyexpert evaluation. After identifying the propositionalrequirements of the methodologiesbeing considered, it maybe appropriateto re-examinethe overall systemwithin the Analysis phase, reconsidering the player frameworks; this can enable the problem to be differently defined, i.e. whether a difference betweenexpressedand perceivedobjectives should be identified, or whetherthe very nature of a given player or player set should be redefmed. New 128 propositionsmaythus develop, anddecision tables already determined may in consequence be redefined; thus newobjective.s tables and a new conflict tableau maydevelop. In terms of game theorythis relates to a redefinitionof the gamethat each player is playing, and evenwhetherit is the same perceived game. 7. Finding Stable ORS Oncea conflict tableau has beengenerated, it is appropriateto reducethe modelto a set of feasible ORSsthat are slructurally stable. This mayoccur throughhumaninspection, or throughexpert system inspection,or by the use of oneof the multivariate decision analysis methods(e.g. Zahedi, 1986). could also, for instance, be accomplished usingthe methodknownas Conflict Analysis[Frase~ Hipel, 1984]. This approachprovidesa useful introduction to the logical qualitative aspects whichmaybe associated with scenario formulation within environmentsunder changeand in whichthere are potentially confrontationalcomponents. A tableau maybe generated to represent the objectiveset of feasible scenariospossibleto each conflict situation. Players normallyhavedistinct preferencesand biases, and under this condition there will be a distinct and different preference ordering of scenarios whichshould be considered whenundertakinga stability analysis. The modelling approach can be extended by makinga secondcircuit of the Choicesphase; this can occur by makingthe conflict tableau the determinantfor establishinga set of futures, since feasible objectives within the ORSwill impactthe properties and goals of each player to a degree which mayor maynot be discemable (fig. 3). Other circuits of the Choices phase might also occurperhapsprior to this. Throughoutthe outcomeevaluation and systems analysis, the futures can also be examined against the goals table which can contribute to an investigation of possible divergence, and thus dynamicstability. It should be realised that the selectionof scenarioswithinoneiteration represents the investigation of structural stability. The outcome of the iteration will be a redefinitionof the propositional synthesis, and the creation bf new decisiontables. Themethodology is intended to be sufficiently robust to enable the changingenvironmentwithin the domain of conflict to be satisfactorily representedthroughexaminingpossible futures. A futures tree of newdecision tables modelsis then created with as manybranchesas there are possible futures. Thesefutures are then re-analysedin a continuediteration of the full cyclecreatinga set of possiblefuture trajectories whichwill best suite the players both individuallyandinteractively. Asreal events progress within the domainof conflict, the inappropriatebranchesare shed. Since a branchis a discrete component of a trajectory, inappropriate trajectories are thus also shed. The evaluation of whether certain scenarios represent stable as well as feasible outcomesis determined by the use of the decimal values generatedby each scenario as already explainedin the previoussection. Theseare flags whichenable a logical investigationof the stable situation to be madeaccording to an algorithmic process, rather than by forced logic alone. Certain scenarios are termed UI’s (Unilateral Improvement)enabling the state conditions of the environmentfor that player. A computer program is available to generate solutions to certain classes of problem. Thewholeapproachis particularly suitable for computersimulation, since futures can be modelled examined,and evaluated moreeasily. 8. Simulationand the ModellingCycle This section of the paper is concernedwith someof architectural needsthat shouldbe consideredwhen designinga simulation systemcapable of helping modellers use the conflict modelling cycle to structure a modellingprocess. In order to establish the modellingcycle within a computersystem it is essential to take into accounta numberof aspects. Theseinclude the use of a decision aid to assist in determining the suitability of a modellingapproachor technique; applying the concepts of knowledgebased systems to guide the modellingprocess; andmonitoringthe systemto identify wherethe modellingprocessis in the modelling cycle, its general progress and conceptual suitability, and howthe modelling processescompareswith reality. 8.1 ModellingDecision Support In order to apply computer techniques to the conflict modelling cycle, it is appropriateto discuss the needsof a computersystemto enable it to be able to determinewhatmodelto select, and howto do so. A Modelling Decision Support System (MDSS)can be thought of to consist of three subsystems, the Information Base subsystem, Databasesubsystem,and the Modellingsubsystem. Thethree subsystemswouldbe linked together by an interface block composedof: the DGMS - is a Dialogue Generation and ManagementSystem whichenables the user to use the systemin a user 129 PLAYERS I I Goals Properties Objectives ]] I" Possible4 Futures Dynamic Stability Power Conflict Tableau for Decision Making/ Bargaining Processes Structural stability Domains fig. 2 Relationship between Goals, Objectives, and Properties in the Decision Table Model OAnalysis ~Formal I OSynthesis definition System of system propositions Decision tables (Sr, r=l,p) I player frameworks of perception I Players Goals I Properti es Obj ec t i ves I stable Scenario set of size k at the 0th pass in Conflict Tableau OChoices Ist cycle 0J scenarios 0! ..... 1 0! 2 k I ... I properties regeneration I Futures decision tables S for each of p players OChoices 2nd cycle P li3 Outcome evaluation II II Systems Analysis II II iteration ¯ Redefinition of propositional synthesis and new decision tables Continued Note: si= s± (iS0, ISI, ..., ~Sp) Futures Fig. 3 Decision Table Analysis i Iiteration 130 oriented way; the SBMSa Strategy Base ManagementSystem connected to a model base subsystem,whichenablemodeJlingstrategies to be collected, offered to a modeller, and after a modellingselection has beenmade,applied; in an intelligent system, it will be able to assess the suitability of strategic modelsaccording to the characteristics of the modellingdomainand the modeller; the MIMS- a Modelling Information Management System part of which is an Information Base ManagementSystem 0BMS), connectedto the navigation process and generates an audit trail or history of the modellingprocess (through the database subsystem). Part of the MIMS is a monitoringsystem (MS)whichenables implicit and explicit evaluationof the modelling processto occur. Suitable modelling DSSenvironmentsmaynot only be able to chart a navigationprocessthrough its associatedMIMS, it will also be able to guide the modellerthroughdistinct levels of modelling process. In terms of managementcontrol, the MIMS can be described as havingstrategic models that will providealternativestrategies andprocesses of modelling,including mediaselection, tactical modelswill help the modeller navigate though a modelling domain,operational models will help modeUers solve a current problem,for instance by directing them from a failed test result to a particular area of test. Thesubsystemcan operate with data fromthe data subsystemto generate real time problemorientated modelsthroughthe system interface. Anextension of this is the explicit provision of a full monitoringand performance evaluationsystemwithin-built advisorthat activates the appropriateassistancerequired. This is system is shownin figure 4. 8.2 Other Consideration Otherconsiderations for the systemrelate to the use of deep reasoning methodsfor the evaluation of qualitative aspects of the system. The implementation of a modelling approach as identified within a metamodelhas an implicit requirementto undertake qualitative reasoning. This requiresa high level of intelligence whichfew systemshaveyet beenable to introduce.Thenature of deepmodelling processesis basicallyqualitative. Someapproachesare logically based, identifying appropriate requirements independent of local contexts. Othersare rule based,usingmeta-rulesto determinewhichsurfacerules of a set (with perhaps contradictory or competingelements) to select. Other approachesuse mathematicalmethods,or a combination of all three. One appr~ch in determiningqualitative evaluations, accordingto Kuipers, has a history whichgoes backto the mid 1970’s.Theinvestigatorsof this area of studytend to focus on descriptions of the deep mechanism, capable of representing incompleteknowledge of a structure andbehaviourwithina process. Afinal considerationthat shall be madehere is that of monitoring.Themonitoringof a modelling progress througha particular modellingprocesses mustbe an implicit feature of a modellingsystem. Thedecision rules are determinedby identifyingthe criteria necessary for makinga decision. In a metamodellingenvironment they are typically determined from the experience of recurrent processes. Whenmonitoring suggests that a modeller has not been successful is modelling a process, then feedbackto the systemdecision maker is requiredin order that a decisionrule relating to the current modellercan be adjusted or a newone introduced.Amonitoringsystemis depictedin fig. 5. 9. Conclusion This paper beganby discussingmetamodelling, thus the provisionof a structured modellingapproach. It examinesconflict processesin a systematicway by initially defininga genericmetamodel, andthus offering a modellingcycle involvingthe phasesof analysis, synthesis, and choice, and relating model outcomes to the system under change. It is essential that a structuredapproachis adoptedwhen modelling processes. In particular, individual approachesand mathematicalformulations become moremeaningfulif they are presented in logical association withina modellingcycle. One approach in modelling conflict was to definedecisiontables relating to a conflict that had beenproperly examinedand defined accordingto a set of criteria. Afterdefiningthe propositionalbase tor the conflict, decision tables are created involvingplayer goals, properties, and objectives. Theobjectives becomethe interactive component of the decision tables whencombinedtogether in a Conflict tableau. Thetableau is prunedto a set of stable scenarios. In the next phase of the cycle, the stable .scenarioslead to possiblefutures for the conflict whichmaybe constructedby a variety of methods, applyingeither soft or hard methods.Soft methods in the context of social conflicts mayrequire discussion with participants combinedwith table inspection, or expert system approaches can be adopted. Alternatively,harder approachescan be used, for example through the estimation of probabilities in Markov Processes. 131 MDSS !:!! :!:!:i:i:!:! :!:! :!:i :i:i:!:! :!:i:!i :i:!:i:i :i:i :!:! :!:! :!:! :i~: i:! :i:i :!:! :i:! :i:! :i:i :i:i :!:i :ii :i:i :i:i :i:i :i:i :i:i: ? ii~:: Model base i::ii subsystemi::i:: !!!! Database :::: subsystem I : Navigation & audit ] :~:~i:!{i~ii~ii~ii~ii~iiii~iiii’{i’~i’{i’iiiii trail in modelling ’::’~’~~’"’ ........ ¯ ::::ii I cycle strategy I Information iilbase subsystemii ~~ User iii! i::i:: i:: fill ] ! iii~odellin~ information ~ iiii Relationship Fig. 4 between MDSS Subsystems Player data input Modelling cycle provision --~ of decision rule options ,I Scenario knowledge opt ion selected yes Feedback modelling manager Monitoring I ¯ I Successful solution encountered no ¯ undetermined Modelling strategy problem Intelligent aid in the selection of yes alternative --modelling strategy I Select & I execute modelling strategy Fig. 5 Modelling system monitor for determining if a selection of another modelling strategy is required The result of this choices modelling approach then feeds back into the analysis phase before iteration occurs again, in order to generate a further phase or set of phases within the process. By establishing the linkage between the methodologies the modelling cycle has been shown to have a continuity in the way that these models can be applied to real situations. Finally, some consideration has been made on how such a modelling process might be established within a computer based system. This rests upon the development of a decision support system with an implicit intelligent knowledge based system able to monitor and examine the modelling process as represented within the modelling cycle. 10. References Burch Jr,J.G., Slater,F.R., Grudnitski, G., 1979. Information Systems: theory and practice, 3rd Ed. John Wiley. Checkland, P., 1981, Systems Thinking Systems Practice. John Wiley & Son, Chichester. Checkland, Scholes, 1990, Soft Systems 132 Methodology in Action. Chichester. John Wiley & Son, Fraser,N.M., Hipel,K.W., 1984, Conflict Analysis: Models and Resolutions. North Holland. ~ Klir, G.,J, Folger,T.A., 1988. Fuzzy Sets, Uncertainty, and Information. Prentice-Hall International. Kuipers,B., 1970. Qualitative Simulation. Artificial Intelligence. 29,289-338. Parzen, E., 1962, Stochastic Francisco: Holden Day. Processes. San Petersen,I.D., 1980. The Dynamic Laws of International Political Games1823-1973. Inst. Political Studies, University of Copenhagen, CopenhagenPolitical Studies. Pugh,D.S., 1984, Organisational Theory: Selected Readings. Penguin. Rubenstein,A.A., Haberstroh,C.l., 1965, Some Theories of Organisation. Richard D. Irwin, Homewood,I1. Saaty,T.L., 1989, Conflict Resolution, Praeger, New York Simon,H.A., 1960, The New Science Management Decision. Harper & Brothers, York. of New Spague Jr., R.H, Watson, H.J.,1986, Decision Support Systems. Prentice Hall. Thom, R., 1975, Structural Morphogenesis, Benjamin. Stability and Yolles,M.I., 1985. Simulating Conflict Using Weibull Games. Simulation in Research and Development. Ed. Javor,A. Elsevier Science Pub. (North Holland). Yolles,M.I., 1987. ModellingConflict with Weibull Games.In Analysing Conflict and its Resolution. Ed Bennet,P.G. Clarendon Press, Oxford. Yolles,M., Price,A, 1989. Generic and Contextual Modelling of Conflict in Knowledge Based Decision Support System. Unpublished Yolles, M.I., 1992a. The Conflict Modelling Cycle. J. Conflict Processes, 1,1. Yolles, M., 1992, "Towards Simulation of the Conflict Modelling Cycle". Chinese Association of Modelling and Simulation conference in Beijing, October 20-23. Yolles, M.I., Kemp, G., 1992. The Dynamics of Culturalism. J. Conflict Processes, 1,1. Zahedi, F., 1986, The Analytic Hierarchy Process. Interfaces 16,pp96-108. Institute of Management Science.