Xinyi LU

Dissertation

Supervisor: Prof. Frank BIOCCA

As romantic AI companions become increasingly popular, questions about how people initiate, maintain and perceive these relationships have gained academic attention. This dissertation aims to adopt a mixed- methods approach to investigate ffve research questions. It suggests that romantic AI companionship can be meaningful but also fragile as a form of lowrisk,emotionally responsive intimacy. The ffndings contribute to literature on human–AI relationships by integrating attachment theory, the CASA paradigm, and self-disclosure perspectives, and they offer practical implications for the ethical and relational design of romantic AI platforms.