Evaluating Agents using Social Choice Theory

TL;DR

This paper introduces a novel evaluation framework called Voting-as-Evaluation (VasE) that applies social choice theory to assess agents across various domains, offering interpretability, robustness, and axiomatic foundations.

Contribution

It proposes a new evaluation method based on voting theory, leveraging social welfare functions to improve robustness and interpretability in agent evaluation.

Findings

VasE outperforms Elo and Nash averaging in robustness.

Discoveries of properties in evaluation data not visible through scores.

Maximal lotteries satisfy key consistency and computational efficiency properties.

Abstract

We argue that many general evaluation problems can be viewed through the lens of voting theory. Each task is interpreted as a separate voter, which requires only ordinal rankings or pairwise comparisons of agents to produce an overall evaluation. By viewing the aggregator as a social welfare function, we are able to leverage centuries of research in social choice theory to derive principled evaluation frameworks with axiomatic foundations. These evaluations are interpretable and flexible, while avoiding many of the problems currently facing cross-task evaluation. We apply this Voting-as-Evaluation (VasE) framework across multiple settings, including reinforcement learning, large language models, and humans. In practice, we observe that VasE can be more robust than popular evaluation frameworks (Elo and Nash averaging), discovers properties in the evaluation data not evident from scores alone, and can predict outcomes better than Elo in a complex seven-player game. We identify one particular approach, maximal lotteries, that satisfies important consistency properties relevant to evaluation, is computationally efficient (polynomial in the size of the evaluation data), and identifies game-theoretic cycles.