Current Projects

The following are descriptions of areas of my current research. Most of these activities relate to one or more of my research interests, and many cut across several of my funded projects

Argumentation-based dialogues between agents

Since much human communication is argumentation-based, that is based on the exchange of reasons for and against believing things, it seems that there are advantages to making communication between agents argumentation based as well.

Along with Mike Wooldridge and Peter McBurney of Liverpool University, and Leila Amgoud of IRIT, I have been developing formal models of argumentation, and evaluating their usefullness for inter-agent communication.

This work is currently funded as part of the ASPIC project.

Selected Publications

McBurney, P. and Parsons, S. Games that agents play: A formal framework for dialogues between autonomous agents. Journal of Logic, Language and Information, 11(3), 315-334.

Parsons, S., Wooldridge, M., and Amgoud, L. Properties and complexity of formal inter-agent dialogues. Journal of Logic and Computation, 13(3), 347-376, 2003.

Parsons, S., Wooldridge, M., and Amgoud, A. On the outcomes of formal inter-agent dialogues, Proceedings of the 2nd International Conference on Autonomous Agents and Multi-Agent Systems, Melbourne, 2003.

Robust auction mechanisms

Auctions are an increasingly popular approach to ecommerce and general resource allocation.

One of the problems in deploying auction is know what kind of auction to use for a particular task, a problem exacerbated by the fact that it seems that different tasks require different auctions. With Steve Phelps and Peter Mcburney from Liverpool University, and Elizabeth Sklar from Columbia University, I have been looking at the use of evolutionary computing as a means of generating auction mechanisms. This approach it possible to customise auctions to specific tasks, and can make the choice of auction robust against exploitation by traders who use the auction.

Another kind of robustness is the ability to recover from errors. Along with Mark Klein at the Sloan School of Management, MIT, I have been looking at approaches to make auctions robust in the face of errors typical of distributed environments (such as message loss).

This work is currently funded under the Tools and techniques for automated mechanism design project.

Selected Publications

Phelps, S., McBurney, P., Parsons, S., and Sklar, E., Co-evolutionary auction mechanism design: a preliminary report, Proceedings of the 4th Workshop on Agent Mediated Electronic Commerce, Bologna, 2002.

Klein, M, Parsons,S. and Rodriguez, J. A bluffer's guide to auctions, Technical Rport, Center for Coordination Science, MIT, 2003.

Qualitative probabilistic models

I have a long standing interest in investigating how probabilistic models can be used in the absence of detailed quantitative data.

One ongoing aspect of this work is exploring the limits of argumentation systems as a mechanism for representing and reasoning with probabilistic information, and investigating the trade-offs between the use of fully qualitative information (as in Wellman's Qualitative Probabilistic Networks (QPNs)), and less qualitative models such as the kappa calculus.

Another aspect of this work is our investgation into the correspondances between logical control models of agents, such as the BDI model, and decision theoretic models like MDPs.

Selected Publications

Parsons, S. Qualitative methods for reasoning under uncertainty, MIT Press, 2001.

McBurney, P. J. and Parsons, S. Representing epistemic uncertainty by means of dialectical argumentation, Annals of Mathematics and Artificial Intelligence, 32(1), 125-169, 2001.

Schut, M., Wooldridge, M. and Parsons, S. Reasoning about intentions in uncertain domains, Proceedings of the 6th European Conference on Symbolic and Quantitative Approaches to Reasoning with Uncertainty, Toulouse, 2001.

Rennoij, S., van der Gaag, L., and Parsons, S. Context-specific sign propagation in qualitative probabilistic networks. Artificial Intelligence, 140, 207-230, 2002.

Parsons, S. On precise and correct qualitative probabilistic reasoning,, International Journal of Approximate Reasoning, (to appear).

Robotics

For someone interested in agent coordination,and reasoning under uncertainty, robots are the perfect platform for experiment.

Since September 2002, I have been involved in a RoboCup Four-legged League team, MetroBots which took part in the first American Open and RoboCup 2003.

I also use robots, this time the lowly Lego Mindstorms platform, in my teaching, and have been studying what students think of this.

Selected Publications

Sklar, E., Parsons, S., and Stone, P. RoboCup in Higher Education: A Preliminary Report, Proceedings of the RoboCup Symposium, Padua, 2003.

Sklar, E. and Parsons, S. RoboCupJunior: a vehicle for enhacing technical literacy, Proceedings of the AAAI Robotics Workshop, Edmonton, 2002.

Datamining

I am working with George Hripsak of the Department of Biomedical Informatics at Columbia University on extracting and reasoning with information from a large database of patient records.

In particular, this work is currently focusing on representing and reasoning with temporal data.