Modeling Classification and Inference Learning

Abstract

Human categorization research is dominated by work in classification learning. The field may be in danger of equating the classification learning paradigm with the more general phenomenon of category learning. This paper compares classifi-cation and inference learning and finds that dif-ferent patterns of behavior emerge depending on which learning mode is engaged. Inference learn-ing tends to focus subjects on the internal struc-ture of each category, while classification learning highlights information that discriminates between the categories. The data suggest that different learning modes lead to the formation of different internal representations. SUSTAIN successfully models inference and classification learning by de-veloping different internal representations for dif-ferent learning modes. Other models do not fare as well.