Socially-Optimal Design of Service Exchange Platforms with Imperfect Monitoring

TL;DR

This paper introduces simple, nonstationary rating protocols for service exchange platforms that ensure socially optimal outcomes despite imperfect monitoring, by tailoring behavior prescriptions based on current and past ratings.

Contribution

It presents the first rating protocols that achieve social optimality under imperfect observation and reporting, using only binary ratings and three prescribed behaviors.

Findings

Protocols are effective despite imperfect monitoring.

They require only binary ratings and three prescribed behaviors.

Protocols are nonstationary, adapting to rating histories.

Abstract

In service exchange platforms, anonymous users exchange services with each other: clients request services and are matched to servers who provide services. Because providing good-quality services requires effort, in any single interaction a server will have no incentive to exert effort and will shirk. We show that if current servers will later become clients and want good-quality services, shirking can be eliminated by rating protocols, which maintain ratings for each user, prescribe behavior in each client-server interaction, and update ratings based on whether observed/reported behavior conforms with prescribed behavior. The rating protocols proposed are the first to achieve social optimum even when observation/reporting is imperfect (quality is incorrectly assessed/reported or reports are lost). The proposed protocols are remarkably simple, requiring only binary ratings and three possible prescribed behaviors. Key to the efficacy of the proposed protocols is that they are nonstationary, and tailor prescriptions to both current and past rating distributions.