如何感知AI对话者:无实体对话式社交机器人拟人化对其印象形成效果影响机制的实验研究

How to Perceive AI Communicators:Experimental Research on the Influence of Anthropomorphism on the Effects of Impression Formation of Bodiless Conversational Social Robots

随着以生成式AI为内核的无实体对话式社交机器人的兴起,使用者如何看待其在人机传播中的主体性成为AI技术改变人类认知的元问题。而主要受拟人化程度影响的人机印象形成效果是回答这一问题的感知基础。带感知智能与感知心灵中介效应检验的拟人化程度(低/高)与人机对话情境(任务型/交际型)2×2组间实验结果表明:拟人化程度对人机印象形成的鲜明度、全面度、好感度均存在正向主效应,且在交际型情境下更能影响印象好感度;同时,感知智能与感知心灵是拟人化影响人机印象效果的两个重要中间机制,且后者在交际型情境下的中介作用更强。这说明对于无实体对话机器人而言,存在与实体社交机器人相反的“反恐怖谷效应”。

With the rise of bodiless conversational social robots based on generative AI as the core,how users view their subjectivity in human-machine communication has become a meta-issue for AI technology to change human cognition. The human-machine impression formation effects, which are mainly influenced by the level of anthropomorphism, are the perceptual basis for answering this question. The results of the 2×2 between-groups experiment on the anthropomorphism levels(low/high) and the human-machine communication contexts(task-oriented/social-oriented) with perceived intelligence and perceived mind mediating effects show that: The level of anthropomorphism has a positive main effect on the intensity, breadth, and valence of human-machine impression, and it can bring greater influence on the valence in social-oriented contexts. In the meantime, perceived intelligence and perceived mind are two important intermediate mechanisms through which anthropomorphism influences the effects of human-machine impression, and the latter has a stronger mediating effect in social-oriented contexts. This indicates that for bodiless conversational social robots, there is an “anti-uncanny valley effect” that is opposite to physical social robots.