The Mind in the Machine: How Theory of Mind Capabilities in Generative AI Agents Impact Impressions
Abstract
As Generative Artificial Intelligence (GenAI) rapidly advances, it is crucial to understand how its human-like capabilities impact interactions with users. This study examines impressions of GenAI demonstrating cognitive and affective facets of theory of mind (ToM) in chat-based interactions . Participants (N = 159) evaluated transcripts between human-human, human-GenAI, or GenAI-GenAI pairs under varying ToM conditions. Preliminary findings reveal that, across all agent types, affective ToM increases perceptions of eeriness but also enhances anthropomorphism and likeability. These mixed effects on impressions highlight the delicate balance between advanced capabilities and natural interactions in GenAI design .
