Contaminated Collaboration: Measuring Gender Bias Transfer in LLM-Assisted Student Writing
A new study published on arXiv reveals that AI writing assistants can transfer gender bias into student essays. Researchers found that students using a gender-biased LLM produced essays with more gender-stereotyped language and career suggestions compared to those using a neutral LLM or no AI assistance. The bias transfer was asymmetric, suppressing agency in essays targeting female profiles while largely leaving male-targeted writing unaffected. This highlights the risk of bias propagation in educational AI tools and calls for fairness-aware design. AI
IMPACT Highlights the risk of bias propagation in AI-assisted writing tools, urging fairness-aware design in educational AI.