Peer Neighborhood Mechanisms: A Framework for Mechanism Generalization

TL;DR

This paper introduces a new class of peer prediction mechanisms that extend incentive compatibility to arbitrary distributions by using neighborhood matching, relaxing previous structural assumptions.

Contribution

It proposes a novel framework that generalizes peer prediction mechanisms to arbitrary distributions through neighborhood matching and provides conditions for incentive compatibility.

Findings

Guarantees incentive compatibility under broad belief update conditions

Extends peer prediction to non-categorical, arbitrary distributions

Introduces neighborhood matching as a key concept

Abstract

Peer prediction incentive mechanisms for crowdsourcing are generally limited to eliciting samples from categorical distributions. Prior work on extending peer prediction to arbitrary distributions has largely relied on assumptions on the structures of the distributions or known properties of the data providers. We introduce a novel class of incentive mechanisms that extend peer prediction mechanisms to arbitrary distributions by replacing the notion of an exact match with a concept of neighborhood matching. We present conditions on the belief updates of the data providers that guarantee incentive-compatibility for rational data providers, and admit a broad class of possible reasonable updates.