Exploring Dimensions of Human Likeness in Conversational AI: Implications for Human Identity Threat and Dehumanization

Abstract

This study aims to identify humanlike traits in conversational AI (CAI) that influence human identity threat and dehumanization and to propose design guidelines to mitigate these effects. An online survey was conducted with 323 participants. Four key perceived dimensions of human likeness in CAI were identified through factor analysis: Self-likeness, Agency, Social Adaptability, and Memory & Communication. Structural equation modeling revealed that Self-likeness heightened both human identity threat and dehumanization, whereas Agency decreased dehumanization. Social Adaptability generally diminished the threat to human identity but intensified the effect of Self-likeness on human identity threat. Additionally, age negatively predicted both human identity threat and dehumanization, suggesting that younger individuals are more susceptible to these psychological effects. By uncovering the psychological structure behind users’ perceptions of CAI anthropomorphism, this study deepens the understanding of the psychological and societal implications associated with humanlike CAI and informs its practical and ethical design.