PASCAL - Pattern Analysis, Statistical Modelling and Computational Learning

Programming Agents as a Means of Capturing Self-Strategy
Michal Chalamish, David Sarne and Sarit Kraus
In: Seventh International Conference on Autonomous Agents and Multi-Agent Systems (AAMAS-2008), May 2008, Estoril, Portugal.


In this paper we report results of an extensive evaluation of people's ability to reproduce the strategies they use in sim- ple real-life settings. Having the ability to reliably capture people's strategies in dierent environments is highly desir- able in Multi-Agent Systems (MAS). However, as trivial and daily as these strategies are, the process is not straightfor- ward and people often have a dierent belief of how they act. We describe our experiments in this area, based on the par- ticipation of a pool of subjects in four dierent games with variable complexity and characteristics. The main measure used for determining the closeness between the two types of strategies used is the level of similarity between the ac- tions taken by the participants and those taken by agents they programmed in identical world states. Our results indi- cate that generally people have the ability to reproduce their game strategies for the class of games we consider. However, this process should be handled carefully as some individu- als tend to exhibit a behavior dierent from the one they program into their agents. The paper evaluates one possible method for enhancing the process of strategy reproduction.

PDF - Requires Adobe Acrobat Reader or other PDF viewer.
EPrint Type:Conference or Workshop Item (Paper)
Project Keyword:Project Keyword UNSPECIFIED
Subjects:User Modelling for Computer Human Interaction
ID Code:5091
Deposited By:Sarit Kraus
Deposited On:24 March 2009