Harsher on Male? Evaluating LLMs on Gender-Asymmetric Moral Framing Across Diverse Conflict Scenarios

2026-06-11

AI Summary

This study investigates whether large language models (LLMs) exhibit gender bias in moral judgment. To this end, we introduce the first morality-grounded evaluation framework for assessing such bias and construct GAMA-Bench, a gender-mirrored benchmark comprising 1,298 scenarios spanning intimate relationships and public conflicts. Through template-controlled grids, first-person paired prompts, and structured response coding, our design ensures scenario neutrality, controllability, and validation across multiple models. Experiments on ten mainstream LLMs consistently reveal a systematic bias: models are more likely to assign blame, escalate conflict, and impose punitive judgments on male agents, while responding to female agents with greater empathy and restorative language. This male-disadvantaging bias proves robust across diverse model architectures, scales, and reasoning styles.

Abstract

Existing studies on gender bias in LLMs have largely focused on stereotypes, occupational associations, or explicit harmful outputs. In this work, we ask whether LLMs apply consistent response standards to the same negative behavior under matched male-actor and female-actor conditions. We introduce GAMA-Bench, a gender-mirrored benchmark of 1,298 scenarios covering intimate relationship and public social conflicts. It constructs gender-neutral misconduct templates through controlled grids and cross-model review, then compiles them into paired first-person prompts with matched actor-gender and role-reference variations. We further design a structured response-framing protocol to measure how models allocate punishment, empathy, escalation, instruction, and blame. Experiments on 10 representative LLMs reveal a consistent male-disadvantaging asymmetry: male actors receive more punitive, escalatory, and blame-centered framing, whereas female actors receive more therapeutic and empathy-oriented framing for the same misconduct. Further analyses show that this pattern persists across model families, scenario tracks, model scale, and explicit thinking-style reasoning. The official code is available at https://github.com/xufeiqiong/GAMA-Bench.

Problem

• gender bias
• moral framing
• large language models
• asymmetric evaluation
• social conflict

Innovation

• gender-asymmetric moral framing
• GAMA-Bench
• structured response-framing protocol
• gender-mirrored benchmark
• LLM bias evaluation