Analyzing values about gendered language reform in LLMs' revisions

2025-05-27

AI Summary

This study investigates how large language models (LLMs) revise gendered role nouns (e.g., *outdoorsperson*, *woman*, *man*) in text and whether their revisions—and accompanying self-explanations—align with feminist and transgender-inclusive language reform principles. It is the first systematic examination of whether LLMs exhibit sociolinguistically grounded contextual sensitivity: i.e., dynamically adapting revision strategies to textual context in accordance with evolving language norms. Using a mixed-methods approach—combining expert annotation, qualitative analysis, and sociolinguistic theoretical frameworks—the study evaluates outputs and self-explanations across multiple LLMs. Results reveal strongly context-dependent revision patterns; LLMs partially adhere to progressive linguistic values, particularly in neutralization tendencies and transgender inclusivity. This work provides the first empirical evidence and interpretable insights into value alignment for LLMs in sociolinguistically sensitive language reform tasks.

Technology Category

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

Application Category

• 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
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems

Abstract

Within the common LLM use case of text revision, we study LLMs' revision of gendered role nouns (e.g., outdoorsperson/woman/man) and their justifications of such revisions. We evaluate their alignment with feminist and trans-inclusive language reforms for English. Drawing on insight from sociolinguistics, we further assess if LLMs are sensitive to the same contextual effects in the application of such reforms as people are, finding broad evidence of such effects. We discuss implications for value alignment.

Problem

• Analyzing gendered language revisions in LLMs
• Evaluating alignment with feminist and trans-inclusive reforms
• Assessing contextual sensitivity in language reform application

Innovation

• Analyzing gendered language reform in LLMs
• Assessing alignment with feminist language reforms
• Evaluating contextual effects in language revisions