A Research Agenda for the Modeling and Simulation of Conflict Systems

1. M&SNet Consortium Meeting 2004 Summer Computer Simulation Conference San Jose, California, July 24-29, 2004 A Research Agenda for the Modeling and Simulation of Conflict Systems Levent Yilmaz M&SNet: Auburn M&S Laboratory Computer Science & Engineering Auburn University, Auburn, AL 36849 http://www.eng.auburn.edu/~yilmaz

2. Goals • to overview the common characteristics,structure, and processes involved in conflict systems. • to present challenges in computational modeling of conflicts. • to outline a comprehensive research framework to facilitate • advancement of the state of the art in modeling of conflicts by gleaning useful information from conflict theory, behavioral sciences, and agent theory and • development of new advanced simulationmethodologies to perceive, conceive, and explore realistic conflict scenarios.

3. Plan • Motivation • Conflict Systems: An Overview • A Research Framework for Simulation-Based Study of Conflicts • Conceptualization: Improving the Use of Conflict Theory as well as Behavioral Sciences in Model Formulation and Design • Realization: Agent-Directed Simulation • Augmenting agent interfaces with domain theories. • Advancing realism with agent-based simulation of conflicts • Agent-supported methodologies for next generation PSEs for conflict analysis and peacekeeping studies. • Conclusions

4. Improving the Modeling, Simulation, and Analysis of Conflicts • Motivation:Political, economic, military as well as terrorist conflicts are the most destructive elements of the modern world. • Yet, there is a lack of simulation-based computational frameworks and PSEs that are able to express realistic assumptions, postulates, theories, and dynamics that are explicitly drawn from conflict theory and behavioral sciences. • Need to view conflicts as complex social interaction systems • to understand the characteristics, parameters, and conditions of various types of conflicts to facilitate their computational modeling. • to expand our horizons in modeling and simulationof conflicts from basic dilemmas to realistic complex conflict processes. • to advance the state of the art of simulation science and methodology to better deal with conflicts.

5. Conflict Analysis, Resolution, and Peacekeeping Studies • Associations: IACM - International Association for Conflict Management http://www.iacm-conflict.org/ • Computer-Aid in Conflict Analysis Conflict Datasetshttp://www.pcr.uu.se/research/UCDP/conflict_dataset_catalog/data_list.htm • AI Methods in Conflict Avoidance and Prevention of Crises and Warshttp://www.oefai.at/oefai/aisoc/peace.html • CASCON:http://web.mit.edu/cascon/ • World Bank - Conflict Prevention and Reconstruction – Conflict Analysis Framework • Conflict Forecasting • http://www-marketing.wharton.upenn.edu /forecast/Conflicts/index.html • Selected Organizations - USA: • CRC- Conflict Research Consortium - University of Colorado • ICAR - Institute for Conflict Analysis & Resolution • IGCC - University of California Institute on Global Conflict and Cooperation • PI - Prevention Institute • USIP - US Institute of Peace For a more comprehensive list of resources on conflict and peace keeping studies http://www.eng.auburn.edu/~yilmaz/ADS-Conflict.htm

6. Toward Simulation-Based Problem Solving Environments for Conflict and Peace-Keeping Studies • Yilmaz L. and T. Ören  (2004). “Enriching Computer-Aided Conflict and Peace Studies with Anticipation and Agent-Directed Simulation,” submitted to Agent 2004 Conference on: Social Dynamics: Interaction, Reflexivity and Emergence. • Yilmaz L. and T. Ören  (2004- in press). “Towards Simulation-Based Problem Solving Environments for Conflict Management in Computational Social Science,” In Proceedings of the Agent2003: Challenges in Social Simulation, • Yilmaz L. (2004). “Advancing the Theory and Methodology of Modeling and Simulation to Explore Understanding and Managing Social Conflicts,” to appear in Modeling & Simulation Magazine. • Ören T. and L. Yilmaz (2004). “Behavioral Anticipation in Agent Simulation,” submitted to 2004 Winter Simulation Conference. • Yilmaz L and T. Ören (2004). “Dynamic Model Updating in Simulation with Multimodels: A Taxonomy and Generic Agent-Based Architecture,” to appear in Proceedings of the SCSC'04.

7. Conflict Systems • Structure: Conflict requires at least two parties or two analytically distinct units or entities with mutually exclusive and/or mutually incompatible values, goals, and objectives. Context Conflict • Dynamics: Conflict requires interaction among parties. As such, conflict relations constitute a fundamental social interaction process. Structure Dynamics • Context is the environment within which the conflict occurs (i.e., task structure, situation, institutions, culture, spatial model of the physical environment) influence the way conflict unfolds toward increasing escalation or de-escalation.

8. Conflict Systems - The Structure Structural Components of Conflict Systems • Parties and their characteristics – their values, motivations; aspirations, objectives; their beliefs about conflicts, including their conceptions of strategy and tactics. • Relations between parties –their attitudes, beliefs, and expectations about one another • The structure of the issue – scope and formulation. • The strategy and tacticsemployed by the parties in the conflict in assessing the one another’s utilities, subjective probabilities; use of promises, rewards, incentives. WIDER CONFLICT ENVIRONMENT Interests, constituents, audiences CONFLICT SYSTEM Third Party Party B Party A Norms, Motives Goals etc. Norms, Motives Goals etc.

9. Plan • Motivation • Conflict Systems: An Overview • A Research Framework for Simulation-Based Study of Conflicts • Conceptualization: Improving the Use of Conflict Theory as well as Behavioral Sciences in Model Formulation and Design • Realization: Agent-Directed Simulation • Augmenting agent interfaces with domain theories. • Advancing realism with agent-based simulation of conflicts • Next generation PSEs and training environments for strategic conflict analysis and peacekeeping studies. • Conclusions

10. A Three-Tier Agenda for Advancing Computational Peace and Conflict Studies (1) Advancing the Computational Modeling of Conflict Systems:Agent theory enables modeling actions, dynamics involved in changing preferences and utility, attitudes, objectives, goal-driven behavior, as well as social guiding principles (i.e., norms, culture). (2) Advancing Behavioral Realism in the Simulation of Conflicts(i.e.,Conflicts entail anticipatory behavior: Human behavior and third party interventions are predicated on the perceptions about the plausible future states ofconflicts). (3) Advanced Simulation Methodologies for Conflict Systems(Conventional methodologies lack the flexibility and adaptivity to appropriately deal with multi-aspect, multi-stage, and uncertain social phenomena).

11. I- A Systemic View for Peace and Conflict Studies (Advancing the Modeling of Conflict Systems)   Conflict Data &Knowledge-Base B A   B uA uB A Deliberation Subsystem A vB DB DA B eB eA vA vB B vA A B dB A dA uA d f uB Operational Interaction Subsystem vB OB vA OA yB yA yB yA environment Extending and Realizing the Peace Science Vision of Isard and Smith (1982) with Agent-Directed Simulation

12. Deliberation and Operational Units for Conflicts   Conflict Data &Knowledge-Base DB DA dA dB vB OB OA vA yB yA e

13. The Role of Perception in Decision-Making during Conflicts   Conflict Data &Knowledge-Base B A B uA uB A DB DA eB eA vB B vA A B dB A dA uA uB vB OB OA vA yB yA e

14. Conflict Theories (i.e., power transition, lateral pressure theories)   Conflict Data &Knowledge-Base B A B uA uB A DB DA eB eA vB B vA A B dB A dA uA uB Operational Interaction Subsystem vB OB vA OA yB yA yB yA e

15. Cognitive Deliberation Subsystem and Its Interface to Operational Level   Conflict Data &Knowledge-Base B A   B uA uB A Deliberation Subsystem A vB DB DA B eB eA vA vB B vA A B dB A dA uA d f uB Operational Interaction Subsystem vB OB vA OA yB yA yB yA e

16. II- Behavioral Realism in the Simulation of Conflict Systems i.e., Simulating Domain Theory Anticipatory Systems (Advanced Conflict Behavioral Methodologies) Theory Sciences Computational Agent - Modeling Theory Supported Simulation Agent Theory Conflict System Agent - Directed i.e., Multisimulation Simulation (Next Generation PSEs) Agent - Based Agent Simulation Simulation Improving Behavioral Realism in the Simulation of Conflicts (adaptivity, pro-activity): • Anticipation is a pervasive factor that surrounds many realistic and interesting intelligent processes embedded in social systems, as well as symbolic systems. Conflict are driven by perception and anticipation.

17. Simulating Anticipatory Systems • Conflicts entail anticipatory behavior: Third party interventions are predicated on the perceptions about the future (undesirable) states ofconflict. • Predictive models for the avoidance of wars and crises are already available: AI Methods in Conflict Avoidance and Prevention of Crises and Wars - http://www.oefai.at/oefai/aisoc/peace.html. Issues:How can anticipatory behavior improve social agent interaction? How do anticipations influence attention? How can perception and forecasting be used to realize anticipatory behavior? What is the role of anticipations on motivations and emotions?

18. CONTEXT Social Interdependence Physical . Structure Env Env. SOCIAL Personality Conflict Style ORIENTATION Orientaton Cognitive Emotional Normative Motivational Orientation Orientation Orientation Orientation PSYCHOLOGICAL ORIENTATION Interpretation Subsystem Intention - Behavior The Social Psychology Factor in Conflicts • Behavioral orientations are based on the uncertainties with respect to social environment, cognitive schemas and attribution mechanisms, the state of emotional arousal, and current moods. • Adapting option preferences and utilities based on emergent cognitive and emotional states would be critical to allocate utilities and payoffs for the action tendencies of agents. • The action tendencies are based on the cognitive expectancies, norm values, trust, and motivational orientations leading to cooperation or competition.

19. III- Using Agents as Simulator Design Metaphors i.e., Simulating Domain Theory Anticipatory Systems (Advanced Conflict Behavioral Methodologies) Theory Sciences Computational Agent - Modeling Theory Supported Simulation Agent Theory Conflict System Agent - Directed i.e., Multisimulation Simulation (Next Generation PSEs) Agent - Based Agent Simulation Simulation Advanced Simulation Methodologies for Next Generation PSTEs: • using agents as simulator and model design metaphors will enable dynamic model and simulation update facilities for multisimulation and multimodel simulators.

20. Dynamic Model and Simulation Updating for Exploring Complex Conflict Phenomena • For most realistic complex phenomena the nature of the problem changes as the simulation unfolds. • Conventional methodologies lack the flexibility and adaptivity to appropriately deal with multi-aspect and multi-stage phenomena. • Needto view available knowledge as being contained in the collection of modeling experiments that become plausible and viable given what is known or learned. • Problem:Run-time switching of models based on • interpretation of emergent, potentially unforeseen conditions • to facilitate dynamic run-time model update and replacement for (simultaneous) experimentation with multiple simulation models.

21. Motivation: Adaptive Experience Management in Strategic Leadership Training with Contingency Models ScenarioLine Scenario: An inspection team under the command of the team of participants is at a weapons storage site in a fictional city. The inspection team discovers that weapons from the site are missing and that a hostile crowd is forming around them. As the inspection team radios for help, the members of the command staff must prepare and launch a rescue operation. Evidence begins to mount that the weapons were stolen by paramilitary troops who are motivating the hostile crowd. As additional paramilitary troops stream into the town, the command staff must overcome a series of obstacles in order to rescue the inspection team without incident or injury – A scenario from (Gordon and Iuppa 2003). Contingency Models • Can we foresee all moves in a conflict? What if the original training scenario did not foresee a trainee decision that can result in civilian casualty? • How can we provide to trainers as much freedom as possible, while assuring that the training goals are achieved by exerting control on the scenario flows?

22. Why and When Dynamic Model/Scenario Update is Needed? • Changing Scenarios: For most realistic social dilemmas, the nature of the problem changes as the simulation unfolds. • Ensembles of Models: Our knowledge about the problem (i.e., conflict) being studied may not be captured by any single model or experiment. • Uncertainty: Adaptivity in simulations and scenarios is necessary to deal with emergent conditions for evolving systems in a flexible manner. • Exploration: As simulations of complex phenomena are used to aid intuition, dynamic run-time simulation composition will help identify strategies that are flexible and adaptive.

23. A type of multimodel: metamorphic model - (e.g., egg, larva, pupa, butterfly) There is a predefined sequence for the alternate models. M M1 M2 Mn

24. Another type of multimodel: multiaspect model - (e.g., ice, water, vapor) More than one alternate model can exist at the same time with possible flows of entities (e.g., mass) between submodels M M1 M2 M3

25. Taxonomy of Multimodels (Submodel Structure)

26. Taxonomy of Multimodels (Submodel Activation Behavior)

27. Toward Multisimulation with Dynamic Simulation Updating • Multisimulation(or multisim, for short) is simulation of several aspects of reality in a study. It includes • simulation with single aspect multimodels, • simulation with multiaspect models, and • simulation with multistage models. • Simulation with sequential multimodels allows computational experimentation with several aspects of reality. • Simulation with multiaspect models (or multiaspect simulation) allows computational experimentation with more than one aspect of reality simultaneously. • Simulation with multistage models allows branching of a simulation study into several simulation studies

28. Simulation Branching with Multisimulation frame 1 (1)Select one of them and ignore others. This approach is similar to many cases in traditional simulation where implicit assumptions are not brought to the users attention. (This alternative is not a good one and definitely is not our choice.) (2)Perform an ordered simulation in breadth or depth first manner with alternative contingency models. (In this case, the user cannot easily and intuitively follow the consequences of alternative simulation studies.) (3)Select multisimulation branching (for all or some of them) and observe (visually or through metrics) behavioral and/or structural developments in simulation studies executed in parallel. CS CS MS AS AS AS AS frame 1.1 frame 1.2 CS CS CS CS MS MS AS AS AS AS AS AS AS AS

29. RESEARCH and COOPERATION OPPORTUNITIES IN CONFLICT SYSTEMS MODELING & SIMULATION at The Auburn Modeling and Simulation Laboratory (AMSL) of the M&SNet Topics*: Conflicts are social phenomena that are worth studying because they affect quality of life everywhere. New advanced simulation techniques may offer proper means to model and explore alternative and unforeseen consequences of conflicts. Positions: Auburn Modeling and Simulation Laboratory is seeking graduate students (M.S. and/or Ph.D. level) and visitors within M&SNet organizations to collaborate on a wide range of methodological, theoretical, and applied simulation modeling problems regarding conflict analysis, resolution, and management. Applicants with interest and knowledge in agent-directed simulation and applications of simulation modeling to social science problems as well as to human behavior and conflicts are encouraged to contact Dr. Levent Yilmaz by email at: yilmaz@eng.auburn.edu. Contact: Dr. Levent Yilmaz http://www.eng.auburn.edu/~yilmaz Auburn Modeling and Simulation Laboratory of the M&SNET Computer Science & Engineering College of Engineering Auburn University Auburn, AL 36849 USA * On these topics, AMSL is already cooperating with Dr. Tuncer Ören of the OC-MISS of M&SNet

30. Thank you for your attention ! Questions ?