CASE: a program for simulation of concept learning

The Learning Research and Development Center of the University of Wisconsin is engaged in a long-term multifacet study of concept learning, supported by the U.S. Office of Education. The Concept Attainment Simulation Experiment (CASE) is the facet of this overall effort which utilizes the technology of computer simultation as a vehicle for obtaining a better understanding of the psychological processes involved in the learning of concepts. The long-range goal is the utilization of the insights thus obtained to improve classroom learning. The study of concept learning has a long history within psychology and has received considerable attention in recent years due in part to the book by Bruner, Goodnow, and Austin which delineated strategies for learning concepts. The experimental materials used by Bruner consisted of a finite universe of objects each of which possessed n dimensions; and each dimension could assume k different values. A classification rule (a concept), consisting of a particular combination of dimension values, partitioned the universe into two mutually exclusive sets. In a typical experiment a subject was shown an object which was an exemplar of the set defined by the concept and told his task was to ascertain the classification rule. In order to attain the concept the subject chose objects from the universe and the experimenter indicated the set membership of the object chosen. The object selection-designation procedure continued until the subject could verbalize the correct classification rule and hence the concept had been attained. The experimental situation, the problem to be solved, and learning procedure involved appear reasonably simple and a number of persons have written programs to simulate this type learning experiment---Hunt and Hovland, Hunt, Allen, Wickelgren, and Baker. The book by Hunt provides an excellent review of much of the psychological literature relevant to concept learning as well as a discussion of his own simulation program. Unfortunately the existing programs leave one with the disquieting feeling that although they attain concepts, little has been added to our understanding of the psychological processes involved in concept learning. Most of these programs are at best watered-down algorithms and involve very little of psychological importance. Because of the shortcomings of the existing simulation programs a project was initiated to develop a program which hopefully will eventuate in something of psychological significance.