Political Compass or Spinning Arrow? Towards More Meaningful Evaluations for Values and Opinions in Large Language Models

TL;DR

This paper critiques current evaluation methods for LLMs' values and opinions, advocating for more realistic, unconstrained assessments exemplified by the Political Compass Test, revealing significant variability in model responses.

Contribution

It introduces a more realistic evaluation framework for LLMs' values, highlighting limitations of current multiple-choice methods and proposing open-ended assessments.

Findings

Models respond differently when not forced into multiple-choice format.

Answers vary based on how models are prompted.

Open-ended responses show different answer patterns.

Abstract

Much recent work seeks to evaluate values and opinions in large language models (LLMs) using multiple-choice surveys and questionnaires. Most of this work is motivated by concerns around real-world LLM applications. For example, politically-biased LLMs may subtly influence society when they are used by millions of people. Such real-world concerns, however, stand in stark contrast to the artificiality of current evaluations: real users do not typically ask LLMs survey questions. Motivated by this discrepancy, we challenge the prevailing constrained evaluation paradigm for values and opinions in LLMs and explore more realistic unconstrained evaluations. As a case study, we focus on the popular Political Compass Test (PCT). In a systematic review, we find that most prior work using the PCT forces models to comply with the PCT's multiple-choice format. We show that models give substantively different answers when not forced; that answers change depending on how models are forced; and that answers lack paraphrase robustness. Then, we demonstrate that models give different answers yet again in a more realistic open-ended answer setting. We distill these findings into recommendations and open challenges in evaluating values and opinions in LLMs.