一些大语言模型表现出一致的风险态度
Some Large Language Models Exhibit Consistent Risk Attitudes
浏览论文内容
中文总结 AI 辅助
研究大语言模型在不确定性下的风险态度，引入跨域框架，应用于六个大语言模型和100名人类参与者，发现多数大语言模型有任务内一致性、跨域排序稳定性，且风险态度分布趋向受限，为评估和调整AI系统奠定基础。
中文摘要 AI 辅助
随着人工智能系统在开放式、高风险环境中部署，一个关键维度仍未得到衡量：感知风险如何转化为行动。我们测试大语言模型在不确定性下是否表现出系统且一致的风险态度。我们引入一个跨域框架，将情境风险信念与分类决策解耦，并将其应用于六个代表性大语言模型和100名人类参与者，涉及空间导航、临床分诊和财务分配任务。使用回归模型，我们提取每个智能体的信念到决策映射，并量化风险敏感性和风险态度偏差。我们发现大多数测试的大语言模型表现出：（i）强大的任务内一致性，即在固定任务域内从情境信念到风险决策的稳定映射；（ii）跨域排序稳定性，在不同任务中保持相对风险态势；（iii）相对于更广泛的人类基线，趋向于受限的风险态度分布。这些结果揭示了风险态度是大语言模型行为中一个稳定且先前未被表征的维度，为评估和调整开放式决策中的人工智能系统奠定了基础，并激发了对这些内在行为倾向起源的进一步研究。
英文摘要
As artificial intelligence systems are deployed in open-ended, high-stakes settings, a critical dimension remains unmeasured: how perceived risk is translated into action. We test whether large language models (LLMs) exhibit systematic and consistent risk attitudes under uncertainty. We introduce a cross-domain framework that decouples contextual risk belief from categorical decision, and apply it to six representative LLMs and 100 human participants across spatial navigation, clinical triage, and financial allocation tasks. Using regression models, we extract each agents belief-to-decision mapping and quantify risk sensitivity and risk attitude bias. We find that most tested LLMs exhibit (i) robust intra-task consistency, indicating stable mappings from contextual belief to risk decision within a fixed task domain; (ii) cross-domain rank-order stability, preserving relative risk posture across tasks; and (iii) a convergence toward a restricted risk-attitude distribution relative to the broader human baseline. These results reveal risk attitude as a stable and previously uncharacterized dimension of LLM behavior, establishing a foundation for evaluating and aligning AI systems in open-ended decision-making and motivating further investigation into the origins of these intrinsic behavioral dispositions.
发表机构
- University of Florida（佛罗里达大学）
- Northeastern University（东北大学）
机构由 AI 辅助整理，请以论文原文为准。
