Unequal Voices: How LLMs Construct Constrained Queer Narratives

2025-07-21

AI Summary

This study identifies systematic biases in large language models (LLMs) when generating queer narratives—manifesting as thematic narrowing, pronounced negative valence, and discursive othering—thereby structurally compressing queer lived experiences. Method: We propose a novel tri-dimensional analytical framework—“harmful representation, narrow representation, discursive othering”—to systematically interrogate structural constraints on queer narrative generation in LLMs. Integrating computational text analysis with critical social science methodologies, we design controlled persona-generation experiments to compare representational breadth and complexity across queer and mainstream demographic groups. Contribution/Results: Empirical findings demonstrate that LLMs consistently produce significantly less diverse and less nuanced portrayals of queer individuals than of non-queer counterparts. The study establishes a new paradigm for AI fairness evaluation grounded in intersectional, discourse-aware metrics and advances actionable guidelines for responsible generative AI development.

Technology Category

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

Application Category

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

Abstract

One way social groups are marginalized in discourse is that the narratives told about them often default to a narrow, stereotyped range of topics. In contrast, default groups are allowed the full complexity of human existence. We describe the constrained representations of queer people in LLM generations in terms of harmful representations, narrow representations, and discursive othering and formulate hypotheses to test for these phenomena. Our results show that LLMs are significantly limited in their portrayals of queer personas.

Problem

• LLMs limit queer narratives to stereotypes
• Queer representations lack human complexity
• LLMs exhibit harmful discursive othering

Innovation

• Analyzing LLM queer narrative constraints
• Testing harmful and narrow representations
• Identifying discursive othering in LLMs