Moral Behavior in Social Dilemma Simulation (MoralSim)

该数据集名为MoralSim，旨在评估大型语言模型（LLMs）在具有道德含义的情境下，如何在不同对手行为、道德背景和游戏设置的囚犯困境和公共物品游戏中表现。该数据集包含了32种不同的实验配置，覆盖了不同的道德框架和游戏类型，旨在全面分析LLMs代理在道德行为方面的决策能力，特别是在道德规范与经济激励相冲突的社会困境中。

This dataset, named MoralSim, is designed to evaluate how large language models (LLMs) perform in morally significant scenarios, specifically within prisoner's dilemma and public goods games with varied opponent behaviors, moral contexts, and game settings. Comprising 32 distinct experimental configurations that cover diverse moral frameworks and game types, it aims to comprehensively analyze the decision-making capabilities of LLM agents in terms of moral behavior, particularly in social dilemmas where moral norms conflict with economic incentives.