mlsanityData to 14 Oct 2026

Collective cooperation without individual fidelity in LLM agents

arXiv 2606.30454 ↗First posted 29 Jun 2026physics.soc-ph cs.AIcited by 0

On the map: LLM agents and tool use → Social simulation and multi-agent debate

Abstract (arXiv)

Large language models (LLMs) are increasingly used as agents in simulations of social systems, yet it remains unclear when their behavior can be interpreted as a faithful proxy for human decision-making. Here we test LLM agents against a direct empirical benchmark: a large-scale networked Prisoner's Dilemma experiment with human participants. Using the same interaction protocol, payoff structure, and network topologies, we compare nine open-weight LLMs with the human data. The selected model reproduces several macro-level features of cooperation dynamics, including the early decline and later stabilization of cooperation. This aggregate agreement, however, does not extend uniformly to finer levels of behavior. LLM populations underestimate individual-level heterogeneity and generate conditional cooperation patterns that differ from those observed in humans. Adding a fraction of random agents improves some aspects of micro-level agreement, but does not remove the mismatch in decision rules. These findings reveal a macro--micro dissociation in LLM-based social agents: collective outcomes can appear human-like even when the underlying behavioral distributions and mechanisms are not. They suggest that validating LLM agents as human surrogates requires comparisons across aggregate dynamics, individual heterogeneity, and context-dependent decision rules, rather than outcome-level agreement alone.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Social simulation and multi-agent debate, one of 16 directions in LLM agents and tool use. Builds on 14 of its 27 references that are in the map. First posted on arXiv 29 Jun 2026.

Builds on

14 of 27 references.

Closest ancestors · 1

Same direction, newest first

  1. 2025Generative Exaggeration in LLM Social Agents: Consistency, Bias, and Toxicity1 Jul

Other references · 13

  1. 2023LLaMA: Open and Efficient Foundation Language Modelscited by 14,132
  2. 2025Qwen3 Technical Reportcited by 8,820
  3. 2023Can Large Language Models Be an Alternative to Human Evaluations?cited by 734
  4. 2025SmolLM2: When Smol Goes Big -- Data-Centric Training of a Small Language Modelcited by 335
  5. 2023Playing repeated games with Large Language Modelscited by 217
  6. 2024Emergent social conventions and collective bias in LLM populationscited by 123
  7. 2025Diverse Preference Learning for Capabilities and Alignmentcited by 25
  8. 2024Detecting Mode Collapse in Language Models via Narrationcited by 24
  9. 2025The simulation of judgment in LLMscited by 22
  10. 2025When Large Language Models are Reliable for Judging Empathic Communicationcited by 21
  11. 2026Stop Drawing Scientific Claims from LLM Social Simulations Without Robustness Auditscited by 8
  12. 2023Leveraging Large Language Models for Collective Decision-Makingcited by 6
1 more
  1. 2026The Statistical Signature of LLMscited by 2

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2024Static network structure cannot stabilize cooperation among Large Language Model agents
  2. 2024Shall We Team Up: Exploring Spontaneous Cooperation of Competing LLM Agents
  3. 2026Benchmarking large language model agent societies against human behavioural distributions
  4. 2025Will Systems of LLM Agents Cooperate: An Investigation into a Social Dilemma
  5. 2025Evolution of Cooperation in LLM-Agent Societies: A Preliminary Study Using Different Punishment Strategies
  6. 2024Network Formation and Dynamics Among Multi-LLMs
  7. 2025Group size effects and collective misalignment in LLM multi-agent systems
  8. 2026Payoff scaling shapes cooperation in LLM agents across languages