RATIONALITY IN THE SELECTION TASK - EPISTEMIC UTILITY VERSUS UNCERTAINTY REDUCTION

M. Oaksford and N. Chater (1994) presented a Bayesian analysis of the Wason selection task in which they proposed that people choose cards i n order to maximize expected information gain (EIG) as measured by red uction in uncertainty in the Shannon-Weaver information theory sense. It is argued that the EIG measure is both psychologically implausible and normatively inadequate as a measure of epistemic utility (i.e., th e value of knowledge to the individual). The article is also concerned with the descriptive account of findings in the selection task litera ture offered by Oaksford and Chater. First, it is shown that their ana lysis data reported in the recent article of K. N. Kirby (1994) is uns ound; second, an EIG analysis is presented of the experiments of P. Po llard and J St BT Evans (1983) that provides a strong empirical discon firmation of the theory.