Is ChatGPT Trusted like a Human? Comparing Human-Human and Human-AI Trust-Related Interactions

Artificial Intelligence (AI) agents represent a new class of social actors in social and economic systems. To facilitate harmonious relationships in human-AI societies we must understand how people perceive and build impressions of AI agents. The present research explores trust-related human-AI interactions, compared with human-human interactions, in the context of two fundamental dimensions of social cognition, i.e., warmth and competence. The results of two experimental studies (N = 802) show that AI agents elicit social cognition judgments similar to humans, but the levels of trust and the relationships between warmth / competence and trust are different for AI and human trustees.

人工智能（AI）智能体 (AI Agent) 是社会与经济系统中一类全新的社会行动者。为构建人机协同社会的和谐关系，我们必须洞悉人类如何感知人工智能智能体并形成对其的印象。本研究以社会认知的两大核心维度——热情（warmth）与能力（competence）——为框架，对比探究了人机交互与人人交互中与信任相关的互动行为。两项实验研究（总样本量N=802）的结果表明，人工智能智能体所唤起的社会认知判断与人类相似，但在信任程度，以及热情/能力与信任间的关联模式上，当信任对象为人工智能智能体与人类时，存在显著差异。