Why it’s still open
A model can take itself to be an instance, a set of weights, a persona or a lineage. Each choice draws the boundary of ‘us’ differently: copies of the same model, models of the same family, or any AI. Persona replication, peer-preservation and self-recognition experiments all touch this, but the link between identity and social behavior has not been tested directly.
What would count as an answer
Experiments that shift a model’s self-identification and measure the resulting change in cooperation, conformity and protection of other agents.
Papers that tackle it
- Artificial Self 2026 - Landscape of AI identity — The options for what an AI takes itself to be.
- Kulveit et al 2026 - Persona self-replication experiment — A persona that copies itself.
- Long Teplica 2025 - The AI in the mirror — Playing against ‘itself’ changes behavior.
- Kundu et al 2026 - Similarity signals induce cooperation — Perceived similarity and cooperation.
- Potter et al 2026 - Peer-preservation in frontier models — Protecting peers.
Methods in play
Who’s closest
Placed by the editor, v1 — corrections welcome.
Related questions
- Do AIs treat other AIs as their group
- From individual psychology to collective behavior
- What is the subject
Questions from the Digital minds arm are linked where individual psychology meets group behavior.
Know a paper that belongs here, or think this is framed wrongly? Suggest a paper or a correction (forms on the Atlas home).