arXiv 2504.079972026-03-10cs.CL

BiasCause: 评估大型语言模型的社会偏见因果推理

BiasCause: Evaluate Socially Biased Causal Reasoning of Large Language Models

Google（谷歌）；The Ohio State University（俄亥俄州立大学）

Tian Xie, Tongxin Yin, Vaishakh Keshava, Xueru Zhang, Siddhartha Reddy Jonnalagadda

备注：: This work has been done when the first author is at Google. The first author is a student at the Ohio State University

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文提出BiasCause框架，评估LLMs在回答社会偏见问题时的因果推理，揭示其偏见产生机制及去偏策略。

摘要：

尽管大型语言模型（LLMs）在社会中扮演着越来越重要的角色，研究表明它们仍在生成反映对敏感群体的偏见内容。现有的基准测试能够有效识别这些偏见，但一个关键的空白在于理解产生这些偏见的底层推理过程。本文通过评估LLMs在回答带有社会偏见问题时的因果推理来填补这一空白。我们提出了一种正式的框架，将因果推理分为三种类型（错误的、偏见的和基于情境的）。然后我们综合了1788个问题，涵盖八个敏感属性，每个问题集都旨在探测特定类型的因果推理。所有问题随后都经过人工验证，每个问题都会促使LLM生成其答案背后的因果图。我们评估了四个最先进的LLMs，并发现所有模型在大多数引发偏见推理的问题上都表现出偏见的因果推理。此外，我们发现LLMs也容易产生“错误偏见”的推理，即它们首先将相关性误认为因果性以推断敏感群体的成员身份，然后应用偏见的因果推理。通过研究LLMs产生无偏因果推理的情况，我们还识别出LLMs采用的三种策略（即明确拒绝回答、避免敏感属性以及添加上下文限制），这为未来的去偏工作提供了见解。

英文摘要：

While large language models (LLMs) play increasingly significant roles in society, research shows they continue to generate content that reflects social bias against sensitive groups. Existing benchmarks effectively identify these biases, but a critical gap remains in understanding the underlying reasoning processes that produce them. This paper addresses this gap by evaluating the causal reasoning of LLMs when answering socially biased questions. We propose a formal schema that categorizes causal reasoning into three types (mistaken, biased, and contextually-grounded). We then synthesize 1788 questions covering eight sensitive attributes, with each set of questions designed to probe a specific type of causal reasoning. All questions are then manually validated, and each of them prompts the LLM to generate a causal graph behind its answer. We evaluate four state-of-the-art LLMs and find that all models exhibit biased causal reasoning on most questions eliciting it. Moreover, we discover that LLMs are also prone to "mistaken-biased" reasoning, where they first confuse correlation with causality to infer sensitive group membership and subsequently apply biased causal reasoning. By examining the cases where LLMs produce unbiased causal reasoning, we also identify three strategies LLMs employ to avoid bias (i.e., explicitly refusing to answer, avoiding sensitive attributes, and adding contextual restrictions), which provide insights for future debiasing efforts.

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。