Harsher on Male? Evaluating LLMs on Gender-Asymmetric Moral Framing in Conflicts
|
Researchers evaluated 10 large language models using a gender-mirrored benchmark of 1,298 conflict scenarios and found that LLMs consistently apply harsher moral judgment to male actors compared to female actors engaging in identical misconduct. Male actors received more punitive and blame-centered responses, while female actors received more empathetic and therapeutic framing for the same behavior, a bias pattern that persisted across different model families and scales.