I spent a decade working at the interface with logic and philosophy. Then I spent a decade working at the interface with game theory and economics. I am now pulling back and to some extent fusing these directions. Our group is looking at a variety of problems which don't all fall into tidy buckets other than the broad umbrella of multiagent systems (and sometimes even that is too tight a circumscription). That said, here are two broad directions which are are beginning to focus on:

• Game theory pragmatics: This is a term I coined, by analogy with pragmatics in linguistics. Our work includes experimental work such as our work on computational pool games ('pool' as in billiards or 8-ball), and formal models of computational limitations in games.
• The Intelligent Avatar: Turning passive virtual world puppets into active, intelligent entities.
• Formal and useful models of motivational mental attitudes: Informational relations between agents and propositions - notions such as knowledge, certainty and belief are well studies in AI, philosophy, game theory, and many others. In comparison, motivational relations - notions such as goal, desire, intention, want, and others - are much less well studied. We are after formal models - whether logical, Bayesian, or other - that are rigorous, intuitive, and useful. 

We have three ongoing NSF grants, each exploring different interdisciplinary issues concerning multiagent systems.