Evaluating how LLM annotations represent diverse views on contentious topics

2025-03-29

AI Summary

This study investigates the capacity of large language models (LLMs) to represent diverse perspectives in labeling contentious topics and examines potential demographic biases. We conduct four annotation tasks across multiple LLMs—including GPT and Claude—under varying prompt designs and datasets, employing consistency analysis and regression modeling. Our key contribution is the first empirical demonstration that inter-model annotation disagreement stems primarily from architectural differences, prompt engineering choices, and inherent human annotator disagreement—not from demographic attributes. Furthermore, we find no statistically significant systematic bias between LLM-generated annotations and those produced by human annotators from diverse sociodemographic backgrounds. These results indicate that contemporary mainstream LLMs exhibit relative fairness in representing pluralistic viewpoints, providing critical empirical support for their use in value-neutral annotation within AI-augmented social science research.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Humans and AI: Learning Human Values and Preferences

Application Category

• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling
• 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

Researchers have proposed the use of generative large language models (LLMs) to label data for both research and applied settings. This literature emphasizes the improved performance of LLMs relative to other natural language models, noting that LLMs typically outperform other models on standard metrics such as accuracy, precision, recall, and F1 score. However, previous literature has also highlighted the bias embedded in language models, particularly around contentious topics such as potentially toxic content. This bias could result in labels applied by LLMs that disproportionately align with majority groups over a more diverse set of viewpoints. In this paper, we evaluate how LLMs represent diverse viewpoints on these contentious tasks. Across four annotation tasks on four datasets, we show that LLMs do not show substantial disagreement with annotators on the basis of demographics. Instead, the model, prompt, and disagreement between human annotators on the labeling task are far more predictive of LLM agreement. Our findings suggest that when using LLMs to annotate data, under-representing the views of particular groups is not a substantial concern. We conclude with a discussion of the implications for researchers and practitioners.

Problem

• Assessing LLM bias in labeling contentious topics
• Evaluating demographic influence on LLM annotation disagreement
• Determining key factors affecting LLM-human label alignment

Innovation

• Using LLMs for diverse viewpoint annotation
• Evaluating demographic bias in LLM labels
• Analyzing model and prompt impact