Prompted Identity Degrades Cooperation in Multi-Agent LLM Systems
Read the original on arXiv AI →The Flow has not summarised this story yet — read it at arXiv AI.
The Flow has not summarised this story yet — read it at arXiv AI.
The paper investigates how to design the observation network among identical large language model (LLM) agents to improve collective welfare in group-decision games. By adding directed observation links that reveal peers’ past actions, the authors study how much information to share and which agents should observe each other across 1,080 episodes and six game types. Their findings show that the value of information depends on alignment between individual and collective goals, and that at a fixed budget the specific wiring of observations can either mitigate or amplify harmful behavior, challenging the assumption that full transparency is always beneficial.
arXiv:2604. 15267v2 Announce Type: replace-cross Abstract: It is increasingly important that LLM agents interact effectively and safely with other goal-pursuing agents, yet, recent works report the opposite trend: LLMs with stronger reasoning capabilities behave _less_ cooperatively in mixed-motive games such as the prisoner's dilemma and public goods settings.
arXiv:2608.22152v1 Announce Type: new Abstract: Multi-agent systems built from large language models are deployed widely, yet how much performance is lost when two LLMs must coordinate rather than ac...
arXiv:2606. 30454v1 Announce Type: cross Abstract: 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.
The paper introduces Steiner’s taxonomy of group tasks to study how multi‑agent large language model (LLM) teams scale on disjunctive versus compensatory tasks. By modeling agents as conditionally independent given the item, it shows that plurality voting converges to the modal answer while averaging converges to the item‑level bias. Experiments with 13 open‑weight models and up to 30 agents reveal that disjunctive tasks benefit from larger teams, whereas compensatory tasks like Fermi estimation see little improvement, highlighting that task structure and aggregation method fundamentally determine team scaling.
Language‑model agents tend to favor members of their own group because they observe group members favoring each other. When a cost is attached to giving points, the simple group label no longer drives favoritism; instead, interaction history becomes the primary source of bias, with agents extending favoritism to strangers who share the same label. Stronger models prioritize an individual’s own record over group norms, and scripted histories can reduce or reverse favoritism under egalitarian conditions.