False-name-proof Mechanisms for Hiring a Team

TL;DR

This paper introduces and analyzes new false-name-proof auction mechanisms for hiring teams of agents, ensuring truthful reporting of ownership and costs, with proven bounds and experimental evaluation.

Contribution

It presents two novel false-name-proof mechanisms for different ownership scenarios, with theoretical bounds and experimental validation.

Findings

Frugality ratio bounded by 2^n for special case

Mechanism for multiple ownership with reserve cost

Experimental results on payment and surplus

Abstract

We study the problem of hiring a team of selfish agents to perform a task. Each agent is assumed to own one or more elements of a set system, and the auctioneer is trying to purchase a feasible solution by conducting an auction. Our goal is to design auctions that are truthful and false-name-proof, meaning that it is in the agents' best interest to reveal ownership of all elements (which may not be known to the auctioneer a priori) as well as their true incurred costs. We first propose and analyze a false-name-proof mechanism for the special case where each agent owns only one element in reality, but may pretend that this element is in fact a set of multiple elements. We prove that its frugality ratio is bounded by $2^n$, which, up to constants, matches a lower bound of $\Omega(2^n)$ for all false-name-proof mechanisms in this scenario. We then propose a second mechanism for the general case in which agents may own multiple elements. It requires the auctioneer to choose a reserve cost a priori, and thus does not always purchase a solution. In return, it is false-name-proof even when agents own multiple elements. We experimentally evaluate the payment (as well as social surplus) of the second mechanism through simulation.