Should you use LLMs to simulate opinions? Quality checks for early-stage deliberation

2025-04-11

AI Summary

This work investigates the reliability of large language models (LLMs) in simulating human opinions, aiming to assess their viability as cost-effective early-stage substitutes for human surveys—thereby reducing dependence on expensive, domain-specific manual annotations. To this end, we propose the first three-dimensional quality evaluation framework for LLM-based opinion simulation in early-stage assessment, covering logical consistency, output stability, and alignment with stakeholder expectations. The framework integrates logical constraint verification, multi-round prompt robustness testing, expectation modeling, and crowdsourced human–LLM comparative annotation. Empirical evaluation in AI content moderation reveals that no current mainstream LLM passes all three checks, exposing critical failure modes. Concurrently, we release a publicly available dataset comprising both human- and LLM-generated annotations of policy statements.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior

Application Category

• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

The array of emergent capabilities of large language models (LLMs) has sparked interest in assessing their ability to simulate human opinions in a variety of contexts, potentially serving as surrogates for human subjects in opinion surveys. However, previous evaluations of this capability have depended heavily on costly, domain-specific human survey data, and mixed empirical results about LLM effectiveness create uncertainty for managers about whether investing in this technology is justified in early-stage research. To address these challenges, we introduce a series of quality checks to support early-stage deliberation about the viability of using LLMs for simulating human opinions. These checks emphasize logical constraints, model stability, and alignment with stakeholder expectations of model outputs, thereby reducing dependence on human-generated data in the initial stages of evaluation. We demonstrate the usefulness of the proposed quality control tests in the context of AI-assisted content moderation, an application that both advocates and critics of LLMs' capabilities to simulate human opinion see as a desirable potential use case. None of the tested models passed all quality control checks, revealing several failure modes. We conclude by discussing implications of these failure modes and recommend how organizations can utilize our proposed tests for prompt engineering and in their risk management practices when considering the use of LLMs for opinion simulation. We make our crowdsourced dataset of claims with human and LLM annotations publicly available for future research.

Problem

• Assessing LLMs' ability to simulate human opinions effectively
• Reducing reliance on costly human survey data for evaluation
• Identifying failure modes in LLM opinion simulation for risk management

Innovation

• Introducing quality checks for LLM opinion simulation
• Emphasizing logical constraints and model stability
• Reducing dependence on human-generated survey data