Treating expert knowledge as common sense

Since the expert systems movement of the 1980s and 1990s, AI has had the dream of reproducing expert behavior in specialized domains of knowledge, such as medicine or engineering, by collecting knowledge from human experts. But the first generations of expert systems suffered from two problems -- first, the difficulty of knowledge engineering -- acquiring knowledge from experts in the form of rules. Second, brittleness -- as soon as a problem diverted at all from the precision of the expert knowledge, systems failed catastrophically. We present a new approach for creating domain-specific AI systems, based on treating specialized knowledge with a methodology originally developed for collecting, and reasoning with, Commonsense knowledge from nonexpert users. Advantages of this approach include: Easy knowledge engineering from informants and collection from natural language sources; Semi-automatic construction of ontologies and knowledge bases of assertions; Tolerance of ambiguity, vagueness, redundancy, and contradiction; Joint inference between expert knowledge and general Commonsense background knowledge; Efficient inference, both forward and backward, of plausible assertions.