文章背景与核心概要
传统的博弈论模型通常受限于固定的数学策略空间,难以捕捉现代人工智能和多智能体系统中基于自然语言的复杂交互行为。为了填补这一空白,本文提出了“开放策略独裁者博弈”(Open-Strategy Dictator Game, OSDG)这一全新的博弈论框架。在该框架中,参与者的策略直接表现为可公开查阅的自然语言文档,并由大语言模型(LLM)进行上下文语境的解释与仲裁。
通过对各类多样化策略进行循环赛(round-robin)模拟,并结合softmax均衡频率、优势分析以及对合作相对价值的敏感性测试,研究发现:条件合作策略(即对合作者选择分享、对掠夺者选择夺取)在透明环境中具有极强的演化稳健性并占据主导地位。相比之下,无条件策略(总是分享或总是夺取)则处于弱势。这一成果揭示了当智能体能够相互审查决策机制时,条件合作将成为广泛收益参数下的演化主流。
The Open-Strategy Dictator Game: Cooperation Under Mutual Transparency
Authors: Michael Glass
Date: August 14, 2026
Identifier: arXiv:2608.14913
Subjects: Computer Science and Game Theory (cs.GT); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
开放策略独裁者博弈:相互透明环境下的合作
作者: Michael Glass
日期: 2026年8月14日
标识符: arXiv:2608.14913
学科: 计算机科学与博弈论 (cs.GT);人工智能 (cs.AI);多智能体系统 (cs.MA)
Summary
The Open-Strategy Dictator Game (OSDG) is a novel game-theoretic framework where participants define their strategies as natural-language documents. In this model, the "dictator" decides whether to SHARE or TAKE an endowment based on the explicit strategy of the recipient. These interactions are adjudicated by a Large Language Model (LLM) that interprets the strategies in context. Research findings indicate that conditional cooperation—strategies that share with other cooperators but exploit those who take—is evolutionarily robust and dominant in transparent environments.
摘要 开放策略独裁者博弈(OSDG)是一个新颖的博弈论框架,其中参与者的策略被定义为所有参与者均可见的自然语言文档。在此模型中,“独裁者”根据接收者的明确策略决定是“分享”(SHARE)还是“夺取”(TAKE)禀赋。这些交互由大语言模型(LLM)进行裁决,LLM会在上下文中对双方的策略进行解释。研究结果表明,条件合作(即与其它合作者分享,但对采取夺取行为者进行剥削的策略)在透明环境中具有演化稳健性并占据主导地位。
Abstract
We introduce the Open-Strategy Dictator Game (OSDG), a variant of the classic dictator game in which each player's strategy is a natural-language document visible to all participants. The dictator's decision, to SHARE or TAKE an endowment, may depend on the text of the recipient's strategy. A large language model adjudicates each interaction by interpreting the dictator's strategy in the context of the recipient's. We run round-robin tournaments among diverse strategies and analyze the resulting payoff matrix using softmax equilibrium frequencies, dominance analysis, and sensitivity to the relative value of cooperation. Conditionally cooperative strategies, those that share with cooperators and take from exploiters, consistently dominate, while unconditional strategies (always share or always take) are weakly dominated. The results suggest that in environments where agents can inspect each other's decision procedures, conditional cooperation is evolutionarily robust across a wide range of payoff parameters.
引言与核心内容 我们引入了开放策略独裁者博弈(OSDG),这是经典独裁者博弈的一个变体,其中每个玩家的策略都是对所有参与者可见的自然语言文档。独裁者对禀赋是选择分享(SHARE)还是夺取(TAKE),取决于接收者策略的文本。大语言模型通过在接收者策略的上下文中解释独裁者的策略来裁决每次交互。我们在各种策略之间开展了循环赛(round-robin tournaments),并使用softmax均衡频率、优势分析以及对合作相对价值的敏感性来分析由此产生的收益矩阵。条件合作策略(即与合作者分享并从掠夺者那里夺取的策略)始终占据主导地位,而无条件策略(总是分享或总是夺取)则处于弱势地位。结果表明,在智能体能够互相检查彼此决策程序的环境中,条件合作在广泛的收益参数范围内都具有演化上的稳健性。
Access & Resources
- PDF: View Paper
- HTML: Experimental Version
- Source: TeX Source
- License: Creative Commons Attribution 4.0 International

访问与资源
- PDF: 查看论文
- HTML: 实验版本
- 源码: TeX 源码
- 许可证: 知识共享署名 4.0 国际许可协议
Citation
To cite this work, please use the following DOI: https://doi.org/10.48550/arXiv.2608.14913
引用 如需引用本工作,请使用以下 DOI:https://doi.org/10.48550/arXiv.2608.14913