A Computational Model of Wason's Selection Task

Abstract:
We apply an algorithmic learning model of inquiry to model reasoning carried by experimental subjects in Wason's _Selection Task_ that represents reasoning in the task as computation of a decision tree that supervenes on (partial) semantic representations. We argue that the resulting model improves on previous probabilistic (Bayesian) and pragmatic (Relevance theory) models of the task. In particular, it suggests that subjects' selection could in fact be guided by sophisticated patterns of argumentative reasoning.