Qualitative Reasoning about Physical Systems

Complex machines are used, understood and repaired by people with essentially no formal training in physics and engineering-although the design and manufacture of such machines requires a deep knowledge of these subjects, and advanced mathematical reasoning. How is this possible? The way people understand these devices, the conceptual frameworks or "mental models" which they use, must be different from the differential equation based descriptions of the world taught in school; people's models are acquired informally, are largely qualitative, use causal reasoning, and relate directly to the language we speak every day. They are crude, but successful. This collection of articles, which constituted a special issue of the Journal of Artificial Intelligence, presents the most recent work on qualitative reasoning about the real (physical) world. A common theme of all the contributions is explaining how physical systems work-from heat flow to transistors to digital computation. The explanations are so detailed and exact that they can be used by computer programs to reason about physical work in the same kinds of ways that people do. This rapidly developing area of cognitive science, variously called Qualitative or Naive Physics, is of direct psychological interest and has strong connections to theories of linguistic semantics. But there are also immediate technological applications, with much of this work funded by industry in the expectation of building computer systems which can communicate sensibly with people about physical mechanisms. Daniel G. Bobrow is a Research Fellow in the Intelligent Systems Laboratory, Xerox Palo Alto Research Center, editor-in-chief of the Journal of Artificial Intelligence, and Chair of the Governing Board of the Cognitive Science Society. This book inaugurates the series Computational Models of Cognition and Perception. A Bradford Book.