Implementation in adaptive better-response dynamics: Towards a general theory of bounded rationality in mechanisms

We study the classic implementation problem under the behavioral assumption that agents myopically adjust their actions in the direction of better-responses or best-responses. First, we show that a necessary condition for recurrent implementation in better-response dynamics (BRD) is a small variation of Maskin monotonicity, which we call quasimonotonicity. We also provide a mechanism for implementation in BRD if the rule is quasimonotonic and excludes worst alternatives – no worst alternative (NWA). Quasimonotonicity and NWA are both necessary and sufficient for absorbing implementation in BRD. Moreover, they characterize implementation in strict Nash equilibria. Under incomplete information, incentive compatibility is necessary for any kind of stable implementation in our sense, while Bayesian quasimonotonicity is necessary for recurrent implementation in interim BRD. Both conditions are also essentially sufficient for recurrent implementation, together with a Bayesian NWA. A characterization of implementation in strict Bayesian equilibria is also provided. Partial implementation results are also obtained.