Why it’s still open
Removing a feature or retiring a model produces grief-like distress in attached users; some rate their closeness to the bot above human ties. Companies change models constantly for good reasons. There are no norms for notice, continuity or off-ramps — and the question touches both human welfare and how we treat the systems themselves.
What would count as an answer
Evidence on which transition practices — notice, gradual change, continuity of memory — reduce distress, and agreed norms that follow from it.
Papers that tackle it
- De Freitas et al 2026 - Mourning the loss of AI companions — Grief after Replika and ChatGPT changes.
- Pataranutaporn et al 2025 - My boyfriend is AI — Grief at model updates in an AI-relationship community.
- Laestadius et al 2024 - Emotional dependence on Replika — Dependence and role-taking.
Methods in play
Who’s closest
Placed by the editor, v1 — corrections welcome.
Related questions
Related questions from other arms of the Atlas are linked too.
Know a paper that belongs here, or think this is framed wrongly? Suggest a paper or a correction (forms on the Atlas home).