Abstract
<p>Attachment styles have long been shown to shape how individuals form and maintain interpersonal relationships. Extending this perspective to human-AI interaction, the present study examined whether and how individuals’ interpersonal attachment styles were linked to their human-AI emotion-related behaviors and attachment. We further tested whether this process was mediated by cognitive perceptions related to AI’s emotional capacity (i.e., perceived uniqueness of AI) and moderated by individuals’ cognitive disposition to view non-living entities as human-like (i.e., dispositional anthropomorphism). Participants were a regionally representative sample of 1,406 undergraduate students (Mage = 18.98 years, SD = 1.25; 40% male) recruited from 26 universities in Shanghai, China, using a stratified random sampling procedure. Participants reported on their interpersonal attachment styles, perceived uniqueness of AI, anthropomorphism, and two indicators of emotional engagement with AI: personal disclosure and emotional attachment. Results showed that both interpersonal attachment anxiety and attachment avoidance were positively associated with perceived uniqueness of AI, which was in turn associated with greater personal disclosure and stronger emotional attachment to AI. Anthropomorphism further moderated the association between perceived uniqueness of AI and emotional engagement, such that these associations were stronger among individuals with higher levels of anthropomorphism. Similarly, moderated mediation analyses indicated that the indirect effects of interpersonal attachment orientations on emotional engagement with AI through perceived uniqueness of AI were amplified at higher levels of anthropomorphism. Overall, the findings suggest that human-AI emotional engagement may be shaped by both individuals’ human-human attachment-related orientations and their psychological construal of AI as a human-like, relationally meaningful interaction partner.</p>