A Boolean generalization of the information-gain model can eliminate specific reasoning errors

Abstract

In the Wason selection task, subjects show a tendency towards counter-logical behaviour. Evidence gained from this experiment raises questions about the role that deductive logic plays in human reasoning. A prominent explanation of the effect uses an information-gain model. Rather than reasoning deductively, it is argued that subjects seek to reduce uncertainty. The bias that is observed is seen to stem from maximizing information gain in this adaptively rational way. This theoretical article shows that a Boolean generalization of the information-gain model is potentially considered the normative foundation of reasoning, in which case several inferences traditionally considered errors are found to be valid. The article examines how this affects inferences involving both over-extension of logical implication and overestimation of conjunctive probability.