Rationalizing Discursive Anomalies

Abstract

Sunstein's�Infotopia�offers four reasons for thinking that information-pooling via mechanical aggregation of votes is superior to discursive sharing of opinions. This article focuses on two of them�the Common Knowledge Effect and Group Polarisation�showing that both phenomena might have perfectly good Bayesian explanations. Far from constituting 'errors', both can actually contribute to truth-tracking in ways that cannot be accomplished via mechanical aggregation of votes alone.