The revelation principle fails when the format of each agent's strategy is an action

Abstract: In mechanism design theory, a designer would like to implement a social choice function which specifies her favorite outcome for each possible profile of agents' private types. The revelation principle asserts that if a social choice function can be implemented by a mechanism in equilibrium, then there exists a direct mechanism that can truthfully implement it. This paper aims to propose a failure of the revelation principle. We point out that in any game the format of each agent's strategy is either an informational message or a realistic action, and the action format is very common in many practical cases. The main result is that: For any given social choice function, if the mechanism which implements it has action-format strategies, then "\emph{honest and obedient}" will no longer be the Bayesian Nash equilibrium of the direct mechanism, actually the social choice function can only be implemented "\emph{dishonestly and disobediently}" in Bayesian Nash equilibrium by the direct mechanism. Consequently, the revelation principle fails when the format of each agent's strategy is an action.