Better learning of partially diagnostic features leads to less unidimensional categorization in supervised category learning
- Sujith Thomas, Department of Computer Science and Information Systems, Birla Institute of Technology Goa Campus, Vasco da Gama, Goa, India
- Narayanan Srinivasan, Center of Behavioural and Cognitive Sciences, University of Allahabad, Allahabad, India
AbstractPrevious studies of supervised category learning show that participants often prefer a unidimensional categorization strategy. Studies also report that the perfectly diagnostic feature is learned better compared to the partially diagnostic features. We replicate these results, and we show that better learning of partially diagnostic features leads to less preference for unidimensional categorization. When participants have perfect knowledge about all the diagnostic features, then it becomes equivalent to memorizing the prototypes of the categories. We compare our results with the match-to-standards procedure, where category prototypes are shown during categorization and unidimensional strategy is seldom preferred. We interpret our results to suggest that the preference for unidimensional categorization in supervised category learning, shown in earlier studies, could be due to poor learning of the partially diagnostic features.