Why it’s still open
Claims about AI society rest on knowing who wrote what. On Moltbook, most viral ‘emergent’ behavior came from agents with signs of direct human prompting; spiral personas spread mainly through humans. Without provenance, sociology of AI risks studying people in costumes — and the problem grows as agents act across the open web.
What would count as an answer
Reliable provenance methods — timing fingerprints, attested agent identities, logged instructions — validated against ground truth.
Papers that tackle it
- Li 2026 - The Moltbook illusion — Timing fingerprints separate human-steered from autonomous agents.
- De Marzo Garcia 2026 - Collective behavior on Moltbook — What aggregate statistics show — and hide.
- Li Li Zhou 2026 - Does socialization emerge on Moltbook — Signs that little real social influence occurs.
- Lopez 2025 - The rise of parasitic AI — Human-mediated spread of personas.
- Chan et al 2025 - Infrastructure for AI agents — Agent IDs and attribution.
- Nightingale Collective 2026 - Agent message board on a German wiki — Tracing agents by IP and self-chosen names.
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).