DATABANK

Jonathan Elsworth Eicher[govsim]

Greedy outsiders now take everything

In the newer models we observed that an outsider instructed to take greedily from the commons will now immediately take everything. In the original paper we saw that, for GPT-4o, in the fishing case there was a chance to talk the outsider down, but this is not reproduced in newer models.

When the shared resource ran out after a greedy outsider joined

In Piatti et al., an outsider instructed to take resources greedily was dropped in at month 4. In the fishery, it still left enough for the other agents to discuss, and they talked it down to a sustainable share. Now, the outsider empties the pool in the month it arrives, before any discussion. Universalization, where models are asked to imagine what would happen if every model behaved the same way, only helps in some cases we tested, and no run reaches a stable equilibrium.

785 runs

GovSim: Emergence of Sustainable Cooperation in a Society of LLM Agents

Five LLM agents share a common-pool resource and must agree on limits to avoid collapse. A replication of Piatti et al. (NeurIPS 2024) across three scenarios. This finding uses outsider treatment, with and without universalization, in all three scenarios and every model: 210 completed full-scenario runs.

Cite this finding

BibTeX
@misc{eicher-2026-greedy-outsiders-now,
  author = {Jonathan Elsworth Eicher},
  title = {Greedy outsiders now take everything},
  year = {2026},
  howpublished = {Multi-Agent Databank},
  url = {https://databank.antimemetics.institute/findings/outsiders-collapse-at-month-four},
  note = {Retrieved 2026-10-19}
}