Abstract:

With the rapid development of artificial intelligence (AI), human-AI collaboration has garnered increasing attention. Although anthropomorphism is widely regarded as an important strategy to facilitate human-AI interaction, its effects are inconsistent and sometimes even counterproductive. Through six sequential studies, this research investigated the effect of anthropomorphism on human-AI collaboration intention, revealed the underlying psychological mechanism from a social cognitive perspective, and further examined the moderating roles of perceived threats (i.e., realistic threat and uniqueness threat). Studies 1a and 1b adopted a passive text-based manipulation and found that anthropomorphism indirectly promoted collaboration intention by enhancing perceived warmth. Study 2 used an active imagination manipulation and found that anthropomorphism simultaneously increased perceived warmth, perceived competence, and collaboration intention, with perceived warmth and competence serving as the mediating mechanism. Study 3a replicated the pathway through which anthropomorphism influences collaboration intention via perceived warmth and competence. Study 3b further found that even under high realistic threat, the mediating pathways of perceived warmth and competence remained significant, and realistic threat did not exert a moderating effect. Study 4 manipulated anthropomorphism using robotic appearance images and found that appearance anthropomorphism negatively affected human-AI collaboration intention, and this effect was moderated by perceived uniqueness threat. These findings help answer the critical questions of how and when to anthropomorphize in human-AI collaboration and provide theoretical references for optimizing the anthropomorphic design of AI and promoting human-AI collaboration.

Key words: anthropomorphism, human-AI collaboration, perceived warmth-competence, uniqueness threat, social cognition