Prior-free Collusion-proof Dynamic Mechanisms

TL;DR

This paper introduces prior-free versions of collusion-proof dynamic mechanisms, extending previous prior-dependent designs to settings without prior distribution assumptions, and demonstrates their approximation guarantees.

Contribution

It develops prior-free variants of TU-GUM and NTU-GUM mechanisms, enabling collusion-proof efficiency without relying on prior information.

Findings

Prior-free NTU-GUM achieves a 1.283-approximation to Pareto efficiency.

The mechanisms extend collusion-proof dynamic design to prior-free settings.

The approach applies to repeated single good allocation problems.

Abstract

For a general class of dynamic stochastic multi-player problems, Cs\'oka, Liu, Rodivilov, and Teytelboym (2024) proposed prior-dependent mechanisms. The Guaranteed Utility Mechanism with transfers (TU-GUM) implements efficiency in a Guaranteed Utility Equilibrium (GUE). Its transfer-free variant (NTU-GUM) implements approximate efficiency in {\epsilon}-GUE. In this paper, we define prior-free versions of both TU-GUM and NTU-GUM. As a special case, we believe that the new prior-free NTU-GUM implements a 1.283-approximation to Pareto efficiency for the repeated single good allocation problem in Fikioris, Banerjee, and Tardos (2024).