Testing Principles of Choice in Categorization

Perceptual categorization is arguably a decision-making task. Lab-based, perceptual categorization research tends to ask participants to classify objects into one of a small set of fixed categories. Decision-making research, however, has revealed a number of choice phenomena that emerge as a function of the size and composition of the choice set. In the current work, we explore the possibility that two rational principles of choice may be violated in categorization: independence of irrelevant alternatives (IIA) and regularity. Critically, such violations deviate from predictions of the Luce Choice Rule, which is assumed by the Generalized Context Model (GCM). The empirical framework used to test for these violations is borrowed from the decision-making phenomenon of the attraction effect. We tested two variations of this paradigm, which involves multiple categories. In the first variation (Experiments 1a and 1b), the set of learned categories varied across participants. In the second variation (Experiments 2a and 2b), the test choice sets varied within participants for a common set of learned categories. Both IIA and regularity were violated when the set of learned categories varied across participants, but they were not violated when the choice sets were varied given a fixed set of learned categories. The GCM can account for the violations, but only by adopting response bias parameters that vary across conditions. These data point to emergent phenomena when classifying stimuli into multiple categories.

The online version contains supplementary material available at 10.1007/s42113-024-00232-6.