Care, code, and consequence: Ethical considerations in the co-design of a conversational AI chatbot for grief

Rapid advances in conversational and generative artificial intelligence (AI) outpace existing ethical frameworks, creating unique challenges in emotionally sensitive domains such as grief. This conceptual paper offers a reflective practice-based examination of responsible research and innovation (RRI) by reflecting on the co-design of an AI grief chatbot developed for bereaved older adults. Adopting a hybrid conceptual–reflective stance, we draw on both qualitative data from co-design workshops and collaborative autoethnographic (CAE) to generate analytic insights with an iterative reflexive thematic analysis identifying five ethical tensions that emerged during development: limited user understanding of AI, unpredictable engagement patterns, uncertainty about AI’s role in mental health support, risks of human–AI relational blurring, and the ethical dilemma of not creating grief-specific technologies. As many people already turn to general-purpose conversational AI for emotional support, these insights underscored the need for a harm-reduction approach—designing safer, more transparent, and better-supported tools rather than ignoring this emerging reality. Drawing on co-design with participants and grief professionals, we offer grounded insights and practical guidance for enacting RRI in highly vulnerable, ethically complex settings.