A knowledge-based program for selecting problem-solving paradigms

Knowledge based systems have been built for a wide variety of application domains, but few areas of AI itself have been explored in this manner. We are building an expert system whose expertise is selecting an appropriate problem solving paradigm (PSP) for a task. It is capable of recommending a total of 1095 different variations of problem solving paradigms. We use the phrase problem solving paradigm for any identifiable problem solving process or its definition at the task level in a sense similar to generic tasks. For example, we consider that model-based diagnosis is a PSP. We believe PSP selection has a major impact on problem solving, but has not been adequately addressed to date in AI research. Most problem reformulation methods are targeted to particular PSPs, thus leaving the PSP selection issue virtually untouched. On the other hand, AI research has been accumulating PSPs, usually accompanied by informal statements as to the kind of problems for which they are suitable. In the thesis, we discuss PSP selection as part of problem solving, describe a representation for PSPs, and document the design and implementation of a knowledge based system for selecting PSPs. We believe systems like this will encourage PSP designers to be more precise about the kind of problems for which their PSPs are designed and will assist others in selecting appropriate paradigms. We also report work on PSP comparison and on relevance analysis for PSP selection. (Copies available exclusively from MIT Libraries, Rm. 14-0551, Cambridge, MA 02139-4307. Ph. 617-253-5668; Fax 617-253-1690.)