Towards More Flexible and Common-Sense Reasoning About Beliefs

Some important research problems within reasoning about beliefs are how to deal with incomplete, inaccurate or uncertain beliefs, when to ascribe beliefs and how to make reasoning about beliefs more common-sensical. We present two systems that attack such problems. The first, called CaseMent, uses case-based reasoning to reason about beliefs. It appears that using cases as opposed to roles has advantages with respect to reusability, context-sensitivity of reasoning, belief ascription and introspection. Although t e program is still in an initial stage, the first results from the existing implementation look promising. The second system, ATI’-Meta, addresses the fact that metaphorical descriptions of mental states play an important role in the coherence of mundane discourse (including stories). ATI’-Meta’s reasoning module has advanced prototype implementation. It incorporates powerful facilities for defeasible and uncertain reasoning. Both systems use simulative reasoning and thereby reduce the amount of explicit self-knowledge they need in ascribing, to other agents, beliefs and reasonings imilar to their own.