Contaminated Collaboration: Measuring Gender Bias Transfer in LLM-Assisted Student Writing

The study investigates the transfer of gender bias from LLMs to student-written career plan essays, revealing that biased prompts lead to gender-differentiated language and stereotypical occupational suggestions. In a controlled experiment with 123 participants, essays generated with gender-biased LLM assistance exhibited a significant agentic gap favoring male-target writing, while female-target writing showed suppressed agency. These findings underscore the need for fairness-aware design in AI-assisted educational tools to mitigate bias propagation.