When the Persona RepliesRethinking Parasocial Interaction in the Age of AI-Driven Virtual Humans

Abstract

The concept of parasocial interaction, introduced by Horton and Wohl (1956) to describe the one-sided intimacy television viewers felt toward broadcasters, is now routinely applied to AI-driven virtual humans, digitally rendered figures that recognize individual users, respond in real time, and sustain a stable persona across encounters. This article argues that such application is a conceptual overstretch. We show that the parasocial concept rests on three implicit premises— non-reciprocity, non-responsiveness, and performer-side control of conten—and that AI-driven virtual humans violate all three simultaneously, dissolving the structural asymmetry on which the concept depends. Rather than discarding or further stretching the concept, we propose a reconceptualization: a parasocial-to-social continuum with a new intermediate category, quasi-social interaction, defined as a machine-mediated relational mode that produces genuine behavioural reciprocity while lacking consciousness, autobiographical continuity, and moral agency. We distinguish the three modes along four dimensions and bound the new category against friendship, classical parasociality, and mere tool-use. The article contributes a diagnosis of a broken premise structure, a new conceptual category for responsive media figures, and a set of testable propositions to guide future empirical research on human–virtual human relationships.