KIK 2008: Agent-Based Simulation – From Cognitive Modelling to Engineering Practice, Workshop at KI 2008
VeranstaltungsortTechnische Universität Dresden
Constructing and executing an agent-based model provides many advantages in comparison to traditional simulation approaches. Increased degrees of freedom in model design, possibilities of modelling behaviour in restricted, heterogeneous geographical and temporal context, etc. are only partial aspects of its great potential. Thus with agent-based simulation, means are available to actually test cognitive models of human decision making in an adequately rich, simulated environment. These efforts may finally result in detailed cognitive models that are grounded in reproducible, practical experience, in addition to abstract hypotheses and theories of cognition.
On the other side, in the agent-based simulation community the need for well-defined and reliable cognitive models is growing as complex human behaviour has to be reproduced in real-world applications. Abstract models based on rational optimization, and/or game-theoretic models are not sufficient any more, if complex decision making in urban space has to be modelled for enhanced pedestrian simulations, or if complex dynamics of team work are to be reproduced.
This workshop aims at bringing these two communities together, demonstrating the current state of art in both research directions. We intend to discuss questions such as
- Why does cognitive science need good engineering practice in agent-based simulation?
- Why do real-world applications of agent-based models really need complex agent models?
- How to solve the problem of missing adequate data for empirical validation of such complex agent models?
- How to determine whether a model is cognitively adequate or even "valid"?
- Can agent-based simulations offer a viable path from single-agent cognitive models to multi-agent cognitive models?
We therefore especially invite position papers on the above mentioned questions for discussion at the workshop. We plan to allow for time slots of at least 30 min per presentation to encourage discussion. As a contribution to the state of the art, we welcome papers describing completed research, work in progress reports as well as problem descriptions with respect to the following topics.
- Examples of agent-based models of human behavior in need of cognitive models
- Examples for cognitive theories that might be testable or usable for agent-based simulations
- Validation and evaluation issues in cognitive modeling and agent-based simulation
- Implementation issues
- Methodological aspects of cognitive modeling and agent-based simulation
- Models of human perception, reasoning and action
- Specific areas of application such as strategies of spatial orientation and navigation
Tuesday, September 23, 2008
8:45 Invited talk
More than Cognition: Practical Applications need Psychological Modeling
Prof. Dr. Harald Schaub (IABGmbH)
10:00 Qualitative Social Science Methods for Multi-Agent Models
Tilman A. Schenk, University of Leipzig
10:30 Appropriating Places in Public Spaces - a Multi-Agent Simulation
Sabine Timpf, University of Augsburg
11:00 Simulating Coping Processes in Critical Situations - An Agent based Approach
Régis Newo and Klaus-Dieter Althoff, University of Hildesheim
11:30 Qualitative Validation From Within – Extension of a Simulation Framework and Test
Björn Raupach, KaHo Sint-Lieven and Franziska Klügl, Örebro University
12:00 Distributed versus centralistic action planning for multi-agent-teams in complex,
Andreas Hirschberger and Ute Schmid, University of Bamberg
14:00 Breakout sessions
16:00 Plenum and Finish