Strongly Rationalizable Virtual Implementation in Social Environments With Evidence

Abstract

This article investigates rationalizable implementation in social environments where agents can provide hard evidence on the private information they possess. Specifically, we study necessary and sufficient conditions for virtual implementation using a notion of Rationalizability that captures forward-induction thinking. Our results non-trivially strengthen those from previous contributions to Rationalizable Implementation with dynamic mechanisms. In particular, we prove that, under a mild economic condition, dynamic mechanisms that rely on evidence can strategically distinguish all agents' information types endowed with different type-dependent messages. Thus, in a large class of social environments, if agents' evidence technologies are responsive, i.e., different types are endowed with different evidence, preference interdependence no longer plays a role in determining which class of Social Choice Functions can be rationalizably virtually implemented.