Approximately Optimal Mechanism Design: Motivation, Examples, and Lessons Learned

TL;DR

This paper discusses the concept of approximately optimal mechanisms in auction design, illustrating how they address limitations of traditional methods through case studies on revenue and welfare maximization.

Contribution

It introduces the approximately optimal mechanism design paradigm and applies it to analyze auction complexity and near-optimality in practical scenarios.

Findings

Simpler auctions can often achieve near-optimal revenue.

Low-dimensional bid spaces may suffice for welfare maximization.

Approximate mechanisms provide insights beyond traditional optimal design.

Abstract

Optimal mechanism design enjoys a beautiful and well-developed theory, and also a number of killer applications. Rules of thumb produced by the field influence everything from how governments sell wireless spectrum licenses to how the major search engines auction off online advertising. There are, however, some basic problems for which the traditional optimal mechanism design approach is ill-suited --- either because it makes overly strong assumptions, or because it advocates overly complex designs. The thesis of this paper is that approximately optimal mechanisms allow us to reason about fundamental questions that seem out of reach of the traditional theory. This survey has three main parts. The first part describes the approximately optimal mechanism design paradigm --- how it works, and what we aim to learn by applying it. The second and third parts of the survey cover two case studies, where we instantiate the general design paradigm to investigate two basic questions. In the first example, we consider revenue maximization in a single-item auction with heterogeneous bidders. Our goal is to understand if complexity --- in the sense of detailed distributional knowledge --- is an essential feature of good auctions for this problem, or alternatively if there are simpler auctions that are near-optimal. The second example considers welfare maximization with multiple items. Our goal here is similar in spirit: when is complexity --- in the form of high-dimensional bid spaces --- an essential feature of every auction that guarantees reasonable welfare? Are there interesting cases where low-dimensional bid spaces suffice?