Finding love in algorithms: deciphering the emotional
contexts of close encounters with AI chatbots

Finding love in algorithms: deciphering the emotional
contexts of close encounters with AI chatbots

human–machine communication, generative AI, computational methods, emotion, intimate relationship, multimodal.

AI chatbots are permeating the socio-emotional realms of human life, presenting both benefits and challenges to interpersonal dynamics and

well-being. Despite burgeoning interest in human–AI relationships, the conversational and emotional nuances of real-world, in situ human–AI so-
cial interactions remain underexplored. Through computational analysis of a multimodal dataset with over 35,000 screenshots and posts from r/

replika, we identified seven prevalent types of human–AI social interactions: intimate behavior, mundane interaction, self-disclosure, play and
fantasy, customization, transgression, and communication breakdown, and examined their associations with six basic human emotions. Our
findings suggest the paradox of emotional connection with AI, indicated by the bittersweet emotion in intimate encounters with AI chatbots,
and the elevated fear in uncanny valley moments when AI exhibits semblances of mind in deep self-disclosure. Customization characterizes the
distinctiveness of AI companionship, positively elevating user experiences, whereas transgression and communication breakdown elicit fear
or sadness.

Han Li 1

, Renwen Zhang

https://doi.org/10.1093/jcmc/zmae015

Oxford University Press on behalf of International Communication Association.

July 16, 2024

PERI IRAWAN

PDF

ENGLISH

TEXT