People increasingly turn to AI chatbots for mental well-being. While these systems promise accessible support, clear gaps remain between chatbot capabilities and people’s expectations. This talk presents findings from two studies on the role of large language models in everyday self-care: the first examined how individuals used LLMs and how counsellors perceived this practice, while the second involved chaplains building and critically reflecting on chatbots for mental well-being. Across both studies, professionals pointed to both the promise and current limitations of AI chatbots in supporting psychological needs. I will illustrate these promises and limitations and outline future directions for designing more meaningful AI support.
Joel Wester is a postdoctoral researcher with the Human-Centred Computing group at the University of Copenhagen. Through qualitative and quantitative empirical studies, he investigates people’s experiences of AI-powered interactions and how design shapes user perceptions.