Universal Mechanisms and Moral Preferences in Implementation

This paper reconsiders implementation of social choice functions defined as mapping from states to consequences, where we require the uniqueness of equilibrium outcome at every state. In contrast with the standard models, we construct only mechanisms that are universal, i.e., are free from the detail of the model specification such as the set of states, and allow each agent to have small moral preference. We show that a single mechanism can implement every incentive compatible social choice function. Moral preferences serve not only to eliminate unwanted equilibria but also to make the central planner's information processing simplified as much as possible in ways that each agent will translate her indescribable private signal into the describable characteristic of the socially optimal alternative.