QueerGen: How LLMs Reflect Societal Norms on Gender and Sexuality in Sentence Completion Tasks

2026-01-28

AI Summary

This study investigates how large language models reproduce heteronormative and cisnormative conventions in text generation, producing measurable biases. Through a sentence completion task, the authors systematically compare masked language models (MLMs) and autoregressive language models (ARLMs) in their generation patterns when prompted with queer-identified, non-queer-identified, and unmarked subjects. Representational disparities are quantified across four dimensions: sentiment, evaluative valence, toxicity, and predictive diversity. The findings reveal that MLMs generate more negative and toxic content for queer subjects, while ARLMs partially mitigate this bias—though closed-source ARLMs unexpectedly produce more harmful outputs for unmarked subjects. This work is the first to demonstrate how model architecture and access restrictions jointly shape the distribution of representational harms related to gender and sexual orientation.

Technology Category

• Natural Language Processing: (Large) Language Models
• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Bias, Fairness & Privacy

Application Category

• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

This paper examines how Large Language Models (LLMs) reproduce societal norms, particularly heterocisnormativity, and how these norms translate into measurable biases in their text generations. We investigate whether explicit information about a subject's gender or sexuality influences LLM responses across three subject categories: queer-marked, non-queer-marked, and the normalized"unmarked"category. Representational imbalances are operationalized as measurable differences in English sentence completions across four dimensions: sentiment, regard, toxicity, and prediction diversity. Our findings show that Masked Language Models (MLMs) produce the least favorable sentiment, higher toxicity, and more negative regard for queer-marked subjects. Autoregressive Language Models (ARLMs) partially mitigate these patterns, while closed-access ARLMs tend to produce more harmful outputs for unmarked subjects. Results suggest that LLMs reproduce normative social assumptions, though the form and degree of bias depend strongly on specific model characteristics, which may redistribute, but not eliminate, representational harms.

Problem

• heterocisnormativity
• Large Language Models
• gender and sexuality
• representational bias
• sentence completion

Innovation

• heterocisnormativity
• representational bias
• sentence completion
• large language models
• queer NLP