User-AI intimacy in digital health

AI-enabled technologies have increasingly been used as alternative tools for enhancing self-mental healthcare. While there are divergent views on the existence and implications of the intimacy that users may perceive towards AI applications as a result of interacting with them, the rapid development and prevalent use of AI in digital health underscore the urgent need to understand how users may perceive feelings of intimacy towards AI applications. In this study, we draw on the theory of the psychology of intimacy and a mixed-method analysis of 4,754 user reviews collected from Wysa-an AI-driven digital health application. The paper develops a conceptual model of User-AI intimacy in digital health, built on five interconnected mechanisms - topical content of self-disclosure, experiential outcomes of self-disclosure, emotional attitudes towards AI, cognitive positioning of AI, and AI technological affordances - and on their co-constitutive relationship. The model shows how mechanisms interact to transform intimate interactions into intimate relationships. The paper discusses the temporal and anthropomorphic conditions of the model and explains how it contributes novel insights to intimacy theory and to the development and management of AI healthcare applications in mental health.