arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2603.29693cs.AI
收藏

测量人工智能的元认知

Measuring (some aspects of) the metacognition of AI

Center for Brain Science, RIKEN(日本立命馆大学脑科学研究中心)

Richard Servajean, Philippe Servajean

AI总结:

本文提出使用meta-d'框架和信号检测理论评估AI的元认知能力,通过实验验证其在不同任务和风险下的决策调节能力。

摘要:

一个稳健的决策过程必须考虑不确定性,尤其是在涉及固有风险的选择中。由于人工智能系统越来越多地集成到决策流程中,管理不确定性越来越依赖这些系统的元认知能力,即它们评估自身决策可靠性和调节决策的能力。因此,采用稳健的方法来测量AI的元认知能力至关重要。本文主要是一个方法论贡献,主张采用meta-d'框架作为评估AI元认知敏感性的黄金标准——生成能够区分正确与错误响应的置信度评分。此外,我们提出利用信号检测理论(SDT)来测量AI在不确定性和风险下自发调节决策的能力。为了展示这些心理物理框架的实用价值,我们对三个大型语言模型(LLMs)——GPT-5、DeepSeek-V3.2-Exp和Mistral-Medium-2508进行了两组实验。在第一组实验中,LLMs执行了初步判断后进行置信度评分。在第二组实验中,LLMs仅执行初步判断,而我们操纵了响应相关的风险。一方面,应用meta-d'框架允许我们沿三个轴进行比较:将LLM与最优性比较、在给定任务上比较不同LLM、以及在不同任务上比较同一LLM。另一方面,SDT允许我们评估LLMs在高风险下是否变得更加保守。

英文摘要:

A robust decision-making process must take into account uncertainty, especially when the choice involves inherent risks. Because artificial intelligence (AI) systems are increasingly integrated into decision-making workflows, managing uncertainty relies more and more on the metacognitive capabilities of these systems; i.e, their ability to assess the reliability of and regulate their own decisions. Hence, it is crucial to employ robust methods to measure the metacognitive abilities of AI. This paper is primarily a methodological contribution that highlights a key limitation of commonly used measures of AI metacognitive sensitivity--the ability to generate confidence ratings that distinguish correct from incorrect responses. We then draw attention to the meta-d' framework, a well-established approach from psychology and neuroscience designed to address this limitation. Moreover, we propose to leverage signal detection theory (SDT) to measure the ability of AIs to spontaneously regulate their decisions based on uncertainty and risk. To demonstrate the practical utility of these psychophysical frameworks, we conduct two series of experiments on three large language models (LLMs)--GPT-5, DeepSeek-V3.2-Exp, and Mistral-Medium-2508.

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。

↑