AGENTS-00/ECML-00 Joint Workshop
on
LEARNING AGENTS
June 3, 2000
Barcelona, Spain

Organizing committee members:

• Peter Stone (Co-Chair), AT&T Labs -- Research, pstone@research.att.com
• Sandip Sen (Co-Chair), Univ. of Tulsa, sandip@kolkata.mcs.utulsa.edu
• Hans-Dieter Burkhard, Humboldt University, hdb@informatik.hu-berlin.de
• Jeff Rosenschein, Hebrew University, Israel, jeff@cs.huji.ac.il
• Moshe Tennenholtz, Technion/Stanford University, moshe@robotics.stanford.edu
• Eiji Uchibe, Osaka University, uchibe@er.ams.eng.osaka-u.ac.jp
• Jose Vidal, University of South Carolina, vidal@sc.edu

Machine learning research has matured into an effective tool at the disposition of the designer/developer of intelligent agent based systems. Such agents often work in environments which are at best partially understood and where the domain characteristics or participants change over time. In addition, agents can serve their associated users much more effectively if they are able to capture the unspecified and/or changing preferences of these users.

Another aspect of agent based systems is that they often are situated in a multiagent environment. Agents in such systems have to interact both with associated users and other agents in their environments. Coordination of the activities of multiple agents, whether selfish or cooperative, is essential for the viability of any system in which multiple agents must coexist. Learning and adaptation are invaluable mechanisms by which agents can evolve coordination strategies that meet the demands of the environments and the requirements of individual agents.

The goal of this workshop is to focus on research that will address unique requirements for agents learning and adapting to their environment. Recognizing the applicability and limitations of current machine learning research when applied to situated agents will be of particular relevance to this workshop.

We emphasize three different ways in which machine learning can be used to enhance the performance of an Agent Based System:

• An agent can learn the preferences and changing priorities of associated users.
• An agent can learn about other agents in the environment in order to compete and/or cooperate with them. An agent can learn from other agents, taking advantage of their experiences and incorporating these into its own knowledge base.
• An agent can learn about other regularities in its environment.

• Benefits of adaptive/learning agents over agents with fixed behavior.
• Evaluation of the effectiveness of individual learning strategies (e.g., case-based, explanation-based, inductive), or multistrategy combinations.
• Characterization of learning and adaptation methods in terms of modeling power, communication abilities, knowledge requirement, processing abilities of individual agents.
• Developing learning and adaptation strategies, or reward structures, for environments with cooperative agents, selfish agents, partially cooperative (will cooperate only if individual goals are not sacrificed) and for environments that can contain mixture of these types of agents.
• Analyzing and constructing algorithms that guarantee convergence and stability of group behavior.
• Analyzing effects of knowledge acquisition mechanism on responsiveness of agents or groups to addition/deletion of other agents from the environment.
• Agents learning via passive or non-intrusive observation of user behaviors.
• Agents learning to serve its user better by observing other agents do their job.
• Evolving agent behaviors or co-evolving multiple agents with similar/opposing interests.
• Investigation of teacher-student relationships between the agent and associated user.

• a brief statement of interest (1 page), or
• a complete paper (3000 words maximum) including keywords and authors' complete address

Direct correspondence to:
Peter Stone
AT&T Labs -- Research
Rm. A273
180 Park Ave.
Florham Park, NJ 07932
OFFICE: 973-360-8333
FAX: 973-360-8970
e-mail: pstone@research.att.com
URL: http://www.research.att.com/~pstone

Sandip Sen's homepage