LLMs Reproduce Stereotypes of Sexual and Gender Minorities

2025-01-10

AI Summary

This study systematically evaluates stereotyping and bias against sexual and gender minority (LGBTQ+) individuals in large language models (LLMs). Addressing limitations of traditional binary frameworks, it extends the Stereotype Content Model (SCM) to non-binary gender identities and diverse sexual orientations, establishing a comparable bias assessment paradigm. Employing a dual-method approach—survey-style prompting and generative text production—the study integrates prompt engineering, contrastive textual analysis, human annotation, and large-scale statistical testing of model responses. Empirical results reveal that mainstream LLMs significantly amplify negative stereotypes—including pejorative associations, role stereotyping, and identity erasure—with bias patterns closely mirroring those observed in human cognition. The work identifies representational harm in generative AI and contributes both a theoretical framework and methodological foundation for fairness evaluation and bias mitigation in LLMs.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Philosophy and Ethics of AI: Bias, Fairness & Equity

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Large language models for search
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems

Abstract

A large body of research has found substantial gender bias in NLP systems. Most of this research takes a binary, essentialist view of gender: limiting its variation to the categories _men_ and _women_, conflating gender with sex, and ignoring different sexual identities. But gender and sexuality exist on a spectrum, so in this paper we study the biases of large language models (LLMs) towards sexual and gender minorities beyond binary categories. Grounding our study in a widely used psychological framework -- the Stereotype Content Model -- we demonstrate that English-language survey questions about social perceptions elicit more negative stereotypes of sexual and gender minorities from LLMs, just as they do from humans. We then extend this framework to a more realistic use case: text generation. Our analysis shows that LLMs generate stereotyped representations of sexual and gender minorities in this setting, raising concerns about their capacity to amplify representational harms in creative writing, a widely promoted use case.

Problem

• Large Language Models
• Stereotypes
• LGBTQ+ Bias

Innovation

• Stereotype Model
• Diverse Gender and Sexuality
• Large Language Models (LLM) Bias