arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.04742cs.CYcs.CL
收藏

推理时的推理选择性减少大语言模型中的隐性社会偏见

Inference-Time Reasoning Selectively Reduces Implicit Social Bias in Large Language Models

Cognitive Science Program Department of Psychological & Brain Sciences Indiana University Bloomington(心理与脑科学系认知科学计划印第安纳大学布卢明顿)

Molly Apsel, Michael N. Jones

AI总结:

该研究通过推理增强减少大语言模型中的隐性社会偏见,揭示推理对公平性评估的影响。

摘要:

基于心理学的构建,先前的研究已经识别出大语言模型(LLMs)中显性偏见和隐性偏见之间的区别。尽管许多LLMs在训练后进行对齐和安全程序以避免显性社会偏见的表达,但它们仍然在类似于隐性关联测试(IAT)的间接任务上表现出显著的隐性偏见。最近的研究进一步表明,推理时的推理会损害依赖隐性统计学习的任务的LLM性能。受隐性关联和人类认知中统计学习的理论联系启发,我们研究推理增强的推理如何影响LLM中的隐性偏见。我们发现,在十五种刻板印象主题的IAT风格评估中,某些模型类别在启用推理时显著减少了测量的隐性偏见。这种效果似乎特定于社会偏见领域,因为我们观察到非社会隐性关联没有相应的减少。随着推理在部署的LLMs中越来越多地被默认启用,这些发现表明它可以在某些系统中有意义地改变公平性评估结果,同时也引发了关于对齐程序如何与推理时的推理相互作用以驱动模型类型之间偏见减少变化的问题。更广泛地说,这项工作突显了认知科学和心理学的理论如何通过提供方法论和解释框架来补充AI评估研究,从而揭示模型行为的新见解。

英文摘要:

Drawing on constructs from psychology, prior work has identified a distinction between explicit and implicit bias in large language models (LLMs). While many LLMs undergo post-training alignment and safety procedures to avoid expressions of explicit social bias, they still exhibit significant implicit biases on indirect tasks resembling the Implicit Association Test (IAT). Recent work has further shown that inference-time reasoning can impair LLM performance on tasks that rely on implicit statistical learning. Motivated by a theoretical link between implicit associations and statistical learning in human cognition, we examine how reasoning-enabled inference affects implicit bias in LLMs. We find that enabling reasoning significantly reduces measured implicit bias on an IAT-style evaluation for some model classes across fifteen stereotype topics. This effect appears specific to social bias domains, as we observe no corresponding reduction for non-social implicit associations. As reasoning is increasingly enabled by default in deployed LLMs, these findings suggest that it can meaningfully alter fairness evaluation outcomes in some systems, while also raising questions about how alignment procedures interact with inference-time reasoning to drive variation in bias reduction across model types. More broadly, this work highlights how theory from cognitive science and psychology can complement AI evaluation research by providing methodological and interpretive frameworks that reveal new insights into model behavior.

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。

↑