Handling Contradictory Data with Metareasoning
Donald R. Perlis, Michael R. Anderson · 2002
Our work has involved extending the meta-reasoning capabilities of the Active Logic inference engine (ALMA) to be able to represent a variety of types of contradiction and positive and negative introspection. This involved both theoretical and implementation effort. In particular, we developed techniques for representing, recognizing, and repairing mistakes that arise in the form of contradictions in the evolving belief set of an agent. This had two aspects: general (commonsense) reasoning, and reasoning attendant to natural-language dialogue. This work led to a number of other publications (see below), including a PhD dissertation completed in 2001.