Learning Optimal Contracts: How to Exploit Small Action Spaces

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Bacchiocchi, Francesco, Castiglioni, Matteo, Marchesi, Alberto, Gatti, Nicola
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917687193501696
author Bacchiocchi, Francesco
Castiglioni, Matteo
Marchesi, Alberto
Gatti, Nicola
author_facet Bacchiocchi, Francesco
Castiglioni, Matteo
Marchesi, Alberto
Gatti, Nicola
contents We study principal-agent problems in which a principal commits to an outcome-dependent payment scheme -- called contract -- in order to induce an agent to take a costly, unobservable action leading to favorable outcomes. We consider a generalization of the classical (single-round) version of the problem in which the principal interacts with the agent by committing to contracts over multiple rounds. The principal has no information about the agent, and they have to learn an optimal contract by only observing the outcome realized at each round. We focus on settings in which the size of the agent's action space is small. We design an algorithm that learns an approximately-optimal contract with high probability in a number of rounds polynomial in the size of the outcome space, when the number of actions is constant. Our algorithm solves an open problem by Zhu et al.[2022]. Moreover, it can also be employed to provide a $\tilde{\mathcal{O}}(T^{4/5})$ regret bound in the related online learning setting in which the principal aims at maximizing their cumulative utility, thus considerably improving previously-known regret bounds.
format Preprint
id arxiv_https___arxiv_org_abs_2309_09801
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Learning Optimal Contracts: How to Exploit Small Action Spaces
Bacchiocchi, Francesco
Castiglioni, Matteo
Marchesi, Alberto
Gatti, Nicola
Computer Science and Game Theory
Machine Learning
We study principal-agent problems in which a principal commits to an outcome-dependent payment scheme -- called contract -- in order to induce an agent to take a costly, unobservable action leading to favorable outcomes. We consider a generalization of the classical (single-round) version of the problem in which the principal interacts with the agent by committing to contracts over multiple rounds. The principal has no information about the agent, and they have to learn an optimal contract by only observing the outcome realized at each round. We focus on settings in which the size of the agent's action space is small. We design an algorithm that learns an approximately-optimal contract with high probability in a number of rounds polynomial in the size of the outcome space, when the number of actions is constant. Our algorithm solves an open problem by Zhu et al.[2022]. Moreover, it can also be employed to provide a $\tilde{\mathcal{O}}(T^{4/5})$ regret bound in the related online learning setting in which the principal aims at maximizing their cumulative utility, thus considerably improving previously-known regret bounds.
title Learning Optimal Contracts: How to Exploit Small Action Spaces
topic Computer Science and Game Theory
Machine Learning
url https://arxiv.org/abs/2309.09801