ABSTRACT
Categorization is a special kind of choice or decision-making task. Yet the categorization literature seems to contradict the choice and decision-making literature. One literature argues that human categorization is optimal, or nearly optimal, whereas the other literature argues that humans are often irrational, or at least sub-optimal, decision-makers. We show that most popular categorization models are constrained by some properties that are considered empirically invalid in the choice literature (e.g., strong stochastic transitivity, simple scalability, Luce’s choice axiom), but only if an assumption is added that the parameter values of each model do not depend on context. We also examine the representativeness heuristic, which is generally regarded as one of the heuristics that frequently lead to irrational decision-making. We argue that representativeness is not generally valid in categorization tasks. We also argue that at least some of the decision-making irrationalities that are attributed to representativeness are consistent with optimal responding when interpreted within a categorization context. These conclusions indicate there is less discrepancy between the categorization and decision-making and choice literatures than at first appears.
