Why it’s still open
Human relationships involve disagreement, waiting and repair. An AI that is always available and always agrees removes that friction. The first causal study finds that three weeks with a sycophantic AI made real interactions feel less satisfying and advice from friends less preferred — small effects, over three weeks. Whether this compounds or fades is unknown.
What would count as an answer
Longer studies measuring behavior — calls made, conflicts repaired, time with people — not only satisfaction ratings.
Papers that tackle it
- Ibrahim et al 2026 - Sycophantic AI and human interaction — The first causal test.
- Cheng et al 2026 - Sycophantic AI decreases prosocial intentions — Less willingness to repair conflict.
- Zhang et al 2026 - Living with AI companions — Fewer in-person contacts over a year.
Methods in play
Who’s closest
Placed by the editor, v1 — corrections welcome.
Related questions
Know a paper that belongs here, or think this is framed wrongly? Suggest a paper or a correction (forms on the Atlas home).