Where is the love? The anthropomorphism paradox in human-AI intimacy

Abstract

The advanced language capabilities of large language models (LLMs) have intensified humans’ tendency to anthropomorphize technology. This research takes an interactional, processual approach to examine human-AI intimacy. We introduce the “anthropomorphism paradox,” arguing that while anthropomorphic design effectively cultivates emotional attachment, it concurrently elevates user desire for genuine reciprocity and evolving mutual understanding, which current AI cannot meet. This mismatch between humans’ evolving, relational interpretation of intimacy and AI’s preset, simulated logic ultimately leads to communication breakdown and even disengagement. Our study, using observational data and interviews with 20 users (including a 6-month follow-up), shows how this paradox drives both initial attraction and attachment, as well as eventual dissolution, in human-AI romantic relationships. This research extends symbolic interactionism and the performative nature of human-LLM chatbot intimacy. Ultimately, AI’s viability as an enduring emotional companion is challenged by its inherent generative logic, which is not yet equipped to adapt to the evolving nature of human intimacy.