Reference: Pinheiro, V.; Furtado, V.; Pinheiro da Silva, P.; McGuinness, D. L. Explaining Problem Solver Answers. Technical Report, Knowledge Systems, AI Laboratory, Stanford University., 2005.
Abstract: Knowledge based systems (KBSs) should explain their answers if their users are expected to understand and thus trust the answers. Problem solvers, KBSs implementing problem solving methods (PSMs), should also explain their answers. Few KBS systems, however, are effective at explaining their answers either because they cannot systematically generate explanations or, when they can, their explanation components cannot easily be extended to new kinds of tasks. In this paper we present an approach enabling problem solvers to explain their answers in a systematic way. To generate proofs for their answers, the approach relies on the fact that problem solvers can retrieve and reuse their PSMs. To generate explanations automatically the approach relies on the Inference Web infrastructure. The approach is implemented for a deployed problem solver tool using explanations to train police teams to perform resource allocation for public safety.
Full paper available as pdf.