![[2602.14471] Socially-Weighted Alignment: A Game-Theoretic Framework for Multi-Agent LLM Systems](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.14471] Socially-Weighted Alignment: A Game-Theoretic Framework for Multi-Agent LLM Systems
Summary
The paper presents a game-theoretic framework called Socially-Weighted Alignment (SWA) for managing multi-agent large language model (LLM) systems, addressing the balance between individual and collective performance in shared environments.
Why It Matters
As LLMs are increasingly deployed in collaborative settings, understanding how to align individual agent objectives with group welfare is crucial. This framework offers insights into optimizing performance and stability, which is vital for the future of AI systems operating in multi-agent contexts.
Key Takeaways
- SWA balances individual agent goals with group welfare using a social weight parameter.
- A critical threshold exists above which agents stabilize demand, preventing congestion.
- The proposed framework can be implemented without complex updates or reinforcement learning.
Computer Science > Multiagent Systems arXiv:2602.14471 (cs) [Submitted on 16 Feb 2026] Title:Socially-Weighted Alignment: A Game-Theoretic Framework for Multi-Agent LLM Systems Authors:Furkan Mumcu, Yasin Yilmaz View a PDF of the paper titled Socially-Weighted Alignment: A Game-Theoretic Framework for Multi-Agent LLM Systems, by Furkan Mumcu and 1 other authors View PDF HTML (experimental) Abstract:Deploying large language model (LLM) agents in shared environments introduces a fundamental tension between individual alignment and collective stability: locally rational decisions can impose negative externalities that degrade system-level performance. We propose Socially-Weighted Alignment (SWA), a game-theoretic framework that modifies inference-time decision making by interpolating between an agent's private objective and an estimate of group welfare via a social weight $\lambda\in[0,1]$. In a shared-resource congestion game with $n$ agents and congestion severity $\beta$, we show that SWA induces a critical threshold $\lambda^*=(n-\beta)/(n-1)$ above which agents no longer have marginal incentive to increase demand under overload, yielding a phase transition from persistent congestion to stable operation near capacity. We further provide an inference-time algorithmic instantiation of SWA that does not require parameter updates or multi-agent reinforcement learning, and use a multi-agent simulation to empirically validate the predicted threshold behavior. Subjects: Multiage...