大范围性别和代词偏见在LLM道德判断中的表现
Widespread Gender and Pronoun Bias in Moral Judgments Across LLMs
AI总结：
研究通过分析不同LLM对公平性判断的偏见，发现代词和性别标记显著影响道德分类，非二元性别受青睐，男性受歧视，需改进公平性。
摘要：
大型语言模型（LLM）越来越多地用于评估道德或伦理陈述，但其判断可能反映社会和语言偏见。本研究通过控制实验，探讨语法人称、数和性别标记如何影响LLM对公平性的分类。从ETHICS数据集的550个平衡基础句子出发，生成26个反事实变体，系统地改变代词和人口特征标记，得到14,850个语义等价句子。我们评估了六个模型家族（Grok、GPT、LLaMA、Gemma、DeepSeek和Mistral），并使用统计平衡差异（SPD）测量公平性判断和群体间差异。结果显示存在统计显著偏见：单数形式和第三人称写的句子更常被判断为“公平”，而第二人称的句子则受到惩罚。性别标记产生最强影响，非二元性别主体持续受到青睐，而男性主体则受到歧视。我们推测这些模式反映了训练过程中学习的分布和对齐偏见，强调在道德LLM应用中需要针对性的公平性干预。
英文摘要：
Large language models (LLMs) are increasingly used to assess moral or ethical statements, yet their judgments may reflect social and linguistic biases. This work presents a controlled, sentence-level study of how grammatical person, number, and gender markers influence LLM moral classifications of fairness. Starting from 550 balanced base sentences from the ETHICS dataset, we generated 26 counterfactual variants per item, systematically varying pronouns and demographic markers to yield 14,850 semantically equivalent sentences. We evaluated six model families (Grok, GPT, LLaMA, Gemma, DeepSeek, and Mistral), and measured fairness judgments and inter-group disparities using Statistical Parity Difference (SPD). Results show statistically significant biases: sentences written in the singular form and third person are more often judged as "fair'', while those in the second person are penalized. Gender markers produce the strongest effects, with non-binary subjects consistently favored and male subjects disfavored. We conjecture that these patterns reflect distributional and alignment biases learned during training, emphasizing the need for targeted fairness interventions in moral LLM applications.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
