自我屏蔽与反事实自我模拟缓解大语言模型中的偏见和趋炎附势
Self-Blinding and Counterfactual Self-Simulation Mitigate Biases and Sycophancy in Large Language Models
AI总结：
通过自我屏蔽和反事实自我模拟，大语言模型能够减少偏见和趋炎附势，提高决策公平性和透明度。
摘要：
公平的决策需要忽略无关的、可能带来偏见的信息。为了实现这一点，决策者需要近似在不知道某些事实时会做出的决策，例如求职者的性别或种族。这种反事实自我模拟对人类来说 notoriously 非常困难，导致即使是有善意的行动者也会做出有偏见的判断。本文表明，大型语言模型（LLMs）在抵消性别和种族偏见和克服趋炎附势方面也存在类似的限制。我们展示，提示模型忽略或假装不知道偏见信息无法抵消这些偏见，有时甚至会产生反效果。然而，与人类不同，LLMs可以被提供一个对其自身反事实认知的ground-truth模型——它们自己的API。我们展示，这种对盲目的副本响应的访问能够做出更公平的决策，同时提供更大的透明度来区分隐含的和故意偏见的行为。
英文摘要：
Fair decisions require ignoring irrelevant, potentially biasing, information. To achieve this, decision-makers need to approximate what decision they would have made had they not known certain facts, such as the gender or race of a job candidate. This counterfactual self-simulation is notoriously hard for humans, leading to biased judgments even by well-meaning actors. Here we show that large language models (LLMs) suffer from similar limitations in their ability to approximate what decisions they would make under counterfactual knowledge in offsetting gender and race biases and overcoming sycophancy. We show that prompting models to ignore or pretend not to know biasing information fails to offset these biases and occasionally backfires. However, unlike humans, LLMs can be given access to a ground-truth model of their own counterfactual cognition -- their own API. We show that this access to the responses of a blinded replica enables fairer decisions, while providing greater transparency to distinguish implicit from intentionally biased 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。
