AuraTracer智迹闻
中文

EVENT DOSSIER

How do LLMs Evaluate Perceived Moral Agency? Investigating Moral Decision-Making in Human-Artificial Agents Interactions

2026-09-07 12:00 Models across 2 days 🔥 47.2 heat score
2sources
2days unfolding
47.2heat score
2mentions
SummaryAI generated

In September 2026, an empirical study published on arXiv for the first time compared the differences in moral agency perception between humans and large language models (LLMs) among different embodied autonomous agents (AAs) in intelligent city scenarios. The study involved 190 human participants and various LLMs. The results showed that humans’ perception of AAs as moral agents was significantly lower than their perception of themselves. However, when faced with specific moral dilemmas, LLMs tended to reason based on the severity of harm and situational urgency, rather than making stable evaluations of the agents themselves. This situational sensitivity characteristic was highly consistent with human judges’ assessments. Given that LLMs are increasingly involved in daily moral decision-making, these findings are considered highly relevant.

Related eventsRELATED EVENTS
Key entitiesKEY ENTITIES
LLMsarXiv

Coverage · reports per dayLANGUAGE SPLIT

Entity relations
LLMs × arXiv1

Integrated timelineUNIFIED TIMELINE

  1. 2026-09-04

    How do LLMs Evaluate Perceived Moral Ag…

    本研究首次实证比较了人类与大型语言模型(LLMs)在智能城市场景中,对具有不同具身形态的自主人工代理人(AAs)感知道德代理(PMA)的评价差异。研究团队利用经过验证的 PMA 量表,对 190 名人类参与者及多种 LLMs 进行了测试。…

  2. 2026-09-07

    How do LLMs Evaluate Perceived Moral Ag…

    本研究首次实证比较了人类与大型语言模型(LLMs)在评估具有不同具身形式的自主人工智能(AAs)感知道德代理(PMA)时的差异,涉及 190 名人类参与者及多种 LLMs。结果显示,人类对 AAs 的 PMA 感知显著低于对人类自身的感知…

SignalsSIGNALS

Keyword heat
  • LLMs1
  • arXiv1

All reports (2)SOURCES

A arXiv cs.CL en 2026-09-04 19:56

How do LLMs Evaluate Perceived Moral Agency? Investigating Moral Decision-Making in Human-Artificial Agents Interactions

本研究首次实证比较了人类与大型语言模型(LLMs)在智能城市场景中,对具有不同具身形态的自主人工代理人(AAs)感知道德代理(PMA)的评价差异。研究团队利用经过验证的 PMA 量表,对 190 名人类参与者及多种 LLMs 进行了测试。结果显示,人类对 AAs 的道德代理感知显著低于对人类自身的感知。然而,在面对具体场景中的道德困境时,LLMs 倾向于从情境出发进行推理,优先考量伤害严重程度和情境紧迫性,而非对代理人本身做出稳定评估,这一情境敏感性特征与人类评分者高度一致。

A arXiv cs.AI en 2026-09-07 12:00

How do LLMs Evaluate Perceived Moral Agency? Investigating Moral Decision-Making in Human-Artificial Agents Interactions

本研究首次实证比较了人类与大型语言模型(LLMs)在评估具有不同具身形式的自主人工智能(AAs)感知道德代理(PMA)时的差异,涉及 190 名人类参与者及多种 LLMs。结果显示,人类对 AAs 的 PMA 感知显著低于对人类自身的感知;但在具体情境的道德困境中,LLMs 倾向于依据伤害严重性和情境紧迫性进行推理,而非对主体本身做出稳定评估,这种情境敏感性也与人类评分者一致。鉴于 LLMs 日益参与日常道德决策,该研究结果具有高度相关性。