Auctions Between Regret-Minimizing Agents

TL;DR

This paper investigates how regret-minimizing software agents behave in repeated auctions, revealing that truth-telling is dominant in first-price auctions but not in second-price auctions, based on theoretical and simulation analysis.

Contribution

It provides the first analysis of regret-minimizing agents in auction settings, highlighting strategic differences between auction types.

Findings

In second-price auctions, agents have incentives to misreport valuations.

In first-price auctions, truthful reporting is a dominant strategy.

Theoretical and simulation results support these strategic behaviors.

Abstract

We analyze a scenario in which software agents implemented as regret-minimizing algorithms engage in a repeated auction on behalf of their users. We study first-price and second-price auctions, as well as their generalized versions (e.g., as those used for ad auctions). Using both theoretical analysis and simulations, we show that, surprisingly, in second-price auctions the players have incentives to misreport their true valuations to their own learning agents, while in the first-price auction it is a dominant strategy for all players to truthfully report their valuations to their agents.