Investigating individual writing style as a contributor to gender gaps in science and technology

arXiv:2204.13805v5 Announce Type: replace-cross Abstract: Gender gaps in how scientific work is evaluated are well documented, but their sources remain debated. We ask whether an overlooked factor---the linguistic style of the writing itself---is gendered and consequential. Drawing on a framework that distinguishes informational features (which emphasize facts) from involved features (which emphasize relationships), we analyze single-authored abstracts of academic papers and patents across all fields of science and technology. Women's writing is systematically more involved than men's---richer in relational, audience-oriented features and higher in the balance of involved to informational language---a difference that holds across scientific fields, in collaborative as well as single-authored work, and in a large open-access biomedical corpus, throughout the full text of papers, not only their abstracts. This stylistic signature also shapes how work is received---papers whose abstracts are more involved are cited more by women and less by men, and the association persists even when papers are compared against their most content-similar alternatives, indicating a gendered signal independent of topic. That a near-costless feature of writing---word choice rather than substance---leaves a systematic, gendered trace on who cites scientific work points to a subtle channel through which evaluation bias may persist, in tension with the universalist ideal that ideas be judged independently of their author.

Related

• Investigating Thinking Behaviours of Reasoning-Based Language Models for Social Bias Mitigation
• EquiSumm : A Gender Bias-Aware Framework for Inclusive Tweet Summarization
• Trait-Aware Policy Optimization for Autoregressive Multi-Trait Essay Scoring
• Toward Responsible and Epistemically Grounded Multilingual LLMs for Computational Social Science and Humanities

Source: arXiv cs.CL | 2026-08-17