Proceedings of the Annual Meeting of the Cognitive Science Society

A consistent finding in research on conditional reasoning isthat individuals are more likely to endorse the valid modus po-nens (MP) inference than the equally valid modus tollens (MT)inference. This pattern holds for both abstract task and prob-abilistic task. The existing explanation for this phenomenonwithin a Bayesian framework (e.g., Oaksford & Chater, 2008)accounts for this asymmetry by assuming separate probabil-ity distributions for both MP and MT. We propose a novelexplanation within a computational-level Bayesian account ofreasoning according to which “argumentation is learning”.We show that the asymmetry must appear for certain priorprobability distributions, under the assumption that the condi-tional inference provides the agent with new information thatis integrated into the existing knowledge by minimizing theKullback-Leibler divergence between the posterior and priorprobability distribution. We also show under which conditionswe would expect the opposite pattern, an MT-MP asymmetry.