chris@starbase.mitre.com (Chris Elsaesser) (03/16/91)
Call for participation: IJCAI-91 WORKSHOP ON REASONING IN ADVERSARIAL DOMAINS Sunday, August 25, 1991 Workshop Description: Any domain involving multiple, independent agents who have goals that cannot be fully and concurrently satisfied, where there is some incentive for the agents not to cooperate, is potentially adversarial. Examples include economic or political competition, legal disputes, labor or hostage negotiation, battle management, and zero-sum games. The ability to reason in adversarial domains is one requirement for AI to advance from problem solving on the shop floor to decision making in the executive office. Adversarial reasoning, like single-agent and multi-agent planning, may be viewed as attempting to deal with goal conflicts. One of the contributions of the AAAI-90 Workshop on Reasoning in Adversarial Domains was a characterization of goal conflict on a scale of utility correlation. Techniques for dealing with conflict on the positive end of that scale, in single-agent planning and multi-agent coordination, may be necessary in certain adversarial domains that encompass a significant component of planning (e.g., battle management). But planning techniques alone are not adequate in domains where an adversary is actively persuing inconsistent goals. At the opposite end of the goal-conflict spectrum, the mini-max strategy used in two-player, zero-sum games requires extension in situations where there is no obvious move-countermove protocol. Mini-max also is too limited if finding cooperative solutions might yield greater expected utility. The purpose of this workshop is to explore the requirements of, and techniques for reasoning in adversarial domains when a pure mini-max strategy is not applicable. Of particular interest are techniques that have heretofore not received much attention in the AI literature, such as partial goal achievement or goal relaxation, such as is required for successful negotiation. Reasoning about the intentions and beliefs in adversarial contexts are of interest, with the planning of deceptions a possible topic of discussion. Researchers in the areas of game theory, computer game playing, multi-agent planning, legal reasoning, negotiation, and relevant application research will be invited to discuss issues such as: modeling an agent's beliefs, use of decision-theoretic techniques, protocols for accounting for an adversary, and formal representation of adversarial domains. In addition, we intend to review progress made in adversarial reasoning applications. Requirements for invitation: Please mail 4 copies of a paper or extended abstract to the Workshop chair. Include an e-mail address of the main point of contact. Submissions will be reviewed for relevance to the topic of the workshop. Several participants will be invited to present short briefings on their work or to participate on discussion panels. The committee regrets that there will not be enough time for all participants to make presentations. Deadline dates: -April 30 for the submission of papers/extended abstracts to workshop chair. -June 1 - invitation to participate in workshop sent to successful applicants along with a workshop registration form. Participants must register for both the conference and the workshop -revised paper/extended abstract in by June 30 -workshop registration form and the $US65.00 fee from participants to be received by IJCAI no later than July 15 Publication of results: All the accepted submissions will be duplicated and provided to each participant. It is the policy of IJCAI that workshop notes are not to be cited. Therefore, inclusion of a paper will not count as "published elsewhere" if it is later submitted to a conference or journal. A summary of the workshop will be published in AI Magazine. Location: IJCAI Workshops will be held on the Market campus, University of Technology, Sydney which is about 500 meters from Darling Harbour and accessible by the monorail. Preliminary Agenda: Note: the agenda will be modified to reflect the interests of participants, as indicated by their submissions. 8:30-9:00 Introductory remarks by committee 9:00-10:30 Applications of reasoning in adversarial domains 10:30-10:45 Break 10:45-12:00 Theoretical issues in reasoning about adversarial domains 12:00-1:00 Lunch 1:00-2:30 Negotiation, Trade-offs, and Goal Relaxation 2:30-2:45 Break 2:45-4:00 Modeling beliefs of an adversary 4:00-5:00 Discussion and Wrap-up Organizing Committee: Christopher Elsaesser (chair) Paul E. Lehner AI Technical Center Dept. of Systems Engineering Mail stop W429 5 Central Module The MITRE Corporation George Mason University 7525 Colshire Drive 4400 University Drive McLean, VA 22102-3481 Fairfax, VA 22030-4444 Phone: (703) 883-6563 Phone: (703) 323-4355 elsaesser@starbase.mitre.org plehner@gmuvax2.gmu.edu FAX: (703) 883-6435 Katia Sycara Jeff Rosenschein The Robotics Institute Ross Building Wean Hall Computer Science Department Carnegie Mellon University Hebrew University Pittsburgh, PA 15213-3890 Givat Ram, Jerusalem katia@isl.ri.cmu.edu Israel Phone: 011-972-2-585-353 jeff@humus.huji.ac.il --------------- end ------------