DAI-List Digest Tuesday, 8 December 1992 Issue Number 100 Topics: Re: Theoretical DAI Re: Theoretical DAI CFP for IJCAI-93 Workshop on Conflict Management Query on Distributed Expert Systems Information Wanted on Real-Time Blackboards Administrivia: Please send submissions to DAI-List@mcc.com. Send other requests, such as changes in your e-mail address, to DAI-List-Request@mcc.com. ---------------------------------------------------------------------- Date: Tue, 8 Dec 92 02:33:01 MST From: malyanka@enws293.EAS.ASU.EDU (Raphael Malyankar) Subject: Theoretical DAI *Should* we go for a single type of theory in DAI? There are so many different types of things in DAI; offhand I would say the (very) broad subfields within DAI include, with some overlap, the following: 1. "Frameworks": coordination methods, organizational structures, coordination schemes, techniques for problem decomposition & distribution, etc.; 2. Planning, with and without the involvement of other agents; 3. Behavioral studies of systems as collections of agents; 4. Reasoning about other agents, modeling other agents and the network, etc. Now, as I see it, class 1 is, roughly, an "architectural" issue, in that researchers construct frameworks for cooperation, interaction protocols, etc., so we use verbal and graphical descriptions, structured to one degree or another. Class 2 is, well, planning, so we get structured verbal descriptions (in which I include algorithms & heuristics) plus some calculus. For class 3 we can go in for theories based on statistical mechanics & statistics, & theories based on equations and distributions & real variables, as in macroeconomics. Class 4 is basically KR and reasoning, and we end up with game theory or one kind of logic or another. Now I may have left out something or the other, & some may disagree with the above broad classification, but the point is that I do not think we should expect a common core (beyond the basic definitions), in the sense of a grand unified theory, or even expect the same type of theory to work in all subfields. It may well be possible to model *some* of the above problems mathematically and end up with descriptions of system behavior based on variables (real or invented). I suppose this is the closest we could come to a common core theory or a G.U.T. Raphael Malyankar malyanka@seine.eas.asu.edu ------------------------------ From: dml@uk.ac.hw.cee.ace Date: Tue, 8 Dec 92 12:00 GMT Subject: Re: DAI-List Digest #99 On theories in DAI: The most useful theories are those that a) allow you to describe what you observe b) enable you to predict what happens next. If we are to synthesize practical DAI systems, then theories should at least accomplish this. If theories don't allow us to predict the behaviour of DAI (or even AI....) systems, then what hope for seeing them used in practice? Nobody will commission systems whose performance cannot be measured against a specification. Remember Hal in 2001 ..... Dave Lane ------------------------------ Date: Mon, 7 Dec 92 14:58:53 -0800 From: mklein@atc.boeing.com Subject: CFP for IJCAI-93 Conflict Management Workshop Call For Papers =============== IJCAI-93 Workshop on Computational Models of Conflict Management in Cooperative Problem Solving Description A central aspect of cooperative problem solving by groups is the avoidance, detection and resolution of conflicts among the participants. This is of great theoretical interest in such research areas as distributed artificial intelligence. It is also of considerable practical importance because of the key role conflict management plays in cooperative problem solving, e.g., in concurrent engineering. Work on conflict management has occurred in a variety of settings, including concurrent engineering, multiagent planning & design, AI and Law, distributed AI (including game theory), GDSS (group decision support systems), CSCW (computer-supported cooperative work), software engineering, sociology, organizational science, public policy and international relations. This work thus includes theoretical groundwork, empirical studies and implemented conflict management systems for human and computational agents. Despite widespread interest, however, there have been few opportunities for researchers addressing these issues in different areas to explore commonalities and benefit from the differing insights each have achieved. The goal of this workshop is to facilitate this kind of cross-fertilization process. The workshop will focus on several key themes: * What lessons do empirical studies of conflict management have to offer for the development of computational models? * What are the current theoretical underpinnings for conflict management, and how can they be applied to practical problems? * How can computers support group conflict management with both human and computational participants? What are the benefits and challenges of the different approaches? * What aspects of conflict management are generic and what are domain-specific? Can the same techniques work with human and computational participants? * How do computational models of conflict management fare in real-world social and organizational settings? Through exploring such themes it is hoped the participants will have a better idea about how they can use related work from other areas, and can begin to outline a single general theory of conflict management that works across multiple domains. Workshop Information This single-day workshop is part of the Workshop Program for IJCAI-93 (the International Joint Conference on Artificial Intelligence), which is to be held in Chambery, France between August 29 - September 3, 1993. The exact workshop date has not been decided yet and will be announced when it becomes available. The workshop will consist of four moderated 75 minute sessions, each made up of: * a brief (5-10 minute) moderators' overview of common themes and key issues * four 10 minute presentations (including questions): presenters will be asked to follow a results-oriented format and to address key issues identified by the moderators. * a discussion panel wherein presenters field questions from the audience and each other. This should be focused on shared issues rather than further explanation of the participant's individual work. Workshop participants are also invited to display posters describing their work. Please note that each attendee must have registered for the main conference and is required to pay an additional 300 FF (about $60 US) fee for the workshop. IJCAI has offered to exempt the workshop fee for one student attendee if he or she agrees to be in charge of taking notes for the whole day. Please let me know if you are interested. Submissions Participation is by invitation only, and will be limited to approximately 35 people of whom 16 will be presenters. Those who wish to attend the workshop should submit four copies of a research abstract no more than 5 pages long focusing on the main contribution of their work in preference to general introductory material, literature review, etc. All submissions will be reviewed by researchers working in a related area. Please include a brief abstract, the author's electronic and physical address information, and indicate if you would like to display a poster on your work at the workshop. Electronic submissions will be accepted only if they are in pure ascii or binhexed Macintosh Word/MacWrite format. Submission deadline: March 1, 1993 Notification date: April 1, 1993 Final date for revised papers: June 1, 1993 We expect that revised versions of the best papers from the workshop will be considered for inclusion in an appropriate journal or published collection. Submissions and questions regarding the workshop should be directed to: Mark Klein Boeing Computer Services PO Box 24346, 7L-64 Seattle, WA 98124-0346 USA mklein@atc.boeing.com Voice: (206) 865-3412 Fax: (206) 865-2964 Organizing Committee Steve Easterbrook University of Sussex Easterbrook@cogs.susx.ac.uk Mark Klein Boeing Computer Services mklein@atc.boeing.com Victor Lesser University of Massachusetts lesser@cs.umass.edu Stephen C-Y. Lu University of Illinois lu@kbesrl.me.uiuc.edu Katia P. Sycara Carnegie Mellon University katia@cs.cmu.edu ------------------------------ From: simmons@quark2.aero.org (CHARLES SIMMONS) Subject: Distributed Expert Systems Date: 5 Nov 1992 10:58 -0800 I'm looking for information on communication protocols for distributed expert systems. Are there any de facto standards emerging in this area? Can anyone recommend any articles or books related to this? Thanks! Charles Simmons The Aerospace Corp. (310) 336-3151 2350 East El Segundo Blvd simmons@arecibo.aero.org El Sequndo, CA 90245 Mail Station M8-108 ------------------------------ From: calistri@oracorp.com (Randy Calistri-Yeh) Subject: Info wanted: real-time blackboards Date: Fri, 20 Nov 1992 17:45:36 GMT We are looking for information on any commercial or public-domain blackboard architectures that would be appropriate for real-time signal-processing applications. This will be used in a fielded system, not for prototype development. Also, if anyone has any case studies or anecdotes about actual experiences with blackboard systems, we would be interested in hearing about them. Randy Calistri-Yeh calistri@oracorp.com ORA Corporation (607) 277-2020 301 Dates Dr. Fax: (607) 277-3206 Ithaca, NY 14850-1313 ------------------------------ End of DAI-List Digest Issue #100 *********************************