1992 年 7 巻 3 号 p. 475-486
This paper describes a model of problem solving for ITS (Intelligent Tutoring System). The performance of an ITS depends largely on how well it knows why the student fails to solve problems. When a human could not get a solution to a problem it may be due to one or more of the following three cases. (1) Some knowledge is missing. (2) Some knowledge could not be used. (3) He/she got tired of solving the problem. Most of the conventional models of problem solving for ITS can take care only the first case. In order to cope with these discrepancies, we propose MIPS (Model of Indexing in Problem Solving) based on a consideration of human problem solving. The major characteristics of MIPS are as follows. (1) It captures the process of human problem solving as a composite one consisting of two phases ; problem understanding and execution of solution methods. (2) The concepts of resource is introduced in order to explain the boundedness and nonmonotonicity of human problem solving. (3) The progress of human problem solving capability is represented as mastering the use of knowledge as well as acquiring new knowledge. MIPS enables to design ITS which stresses the importance of problem understanding and mastering the use of solution methods.