随机环境下惩罚对团队成员道德风险行为的影响  

The Impact of Punishment on Team Members Moral Hazard Behavior in a Stochastic Environment

作　　者：平安 张悟移[1] PING An;ZHANG Wuyi(Faculty of Management and Economics,Kunming University of Science and Technology,Kunming 650093,China;School of Statistics and Mathematics,Yunnan University of Finance and Economics,Kunming 650221,China)

机构地区：[1]昆明理工大学管理与经济学院,云南昆明650093 [2]云南财经大学统计与数学学院,云南昆明650221

出　　处：《昆明理工大学学报(自然科学版)》2025年第2期188-202,共15页Journal of Kunming University of Science and Technology(Natural Science)

基　　金：国家自然科学基金重点项目(11931015);国家自然科学基金面上项目(7216403);国家自然科学基金地区项目(12271471,12461005,62462064);中国工程院院地合作重点项目(2020YNZH);云南省基础研究计划项目(202301AU070084,202301AT070419).

摘　　要：团队是当今组织的核心工作特征,通过惩罚等机制提高团队成员的合作意愿是提高团队绩效的主要手段.但现实环境中个体会综合考虑自身的综合收益进行决策.由于信息不对称难以消除,惩罚等机制有可能产生反效果.基于公共物品博弈和多智能体强化学习方法,构建了智能体团队模型以及可部分完成且存在隐蔽性的任务模型,设置了分别以结果、行为和任务余量为导向的三种惩罚机制,分析了三种惩罚机制在随机环境下对团队成员合作意愿以及隐藏结果的道德风险行为的影响.通过仿真模拟,研究发现惩罚能够加强奖励对合作促进的作用,且以任务余量为导向的惩罚能够有效抑制道德风险.Team is a core characteristic of work in modern organizations,and enhancing the willingness of team members to cooperate through mechanisms such as punishment is a primary means of improving team performance.However,individuals make decisions in real-world environments by considering their overall benefits.Due to the difficulty of eliminating information asymmetry,mechanisms like punishment may have counterproductive effects.Based on the public goods game and multi-agent reinforcement learning methods,a model of an agent team and a task model that can be partially completed with concealment were constructed.Three punishment mechanisms oriented towards outcomes,actions,and task surplus were set up,and the impact of these three mechanisms on the cooperation willingness of team members and the moral hazard behaviors of concealing results and actions in a random environment were analyzed.Simulations were conducted,and the findings suggest that punishment can strengthen the role of rewards in promoting cooperation,and punishment oriented towards task surplus can effectively suppress moral hazard.

关 键 词：惩罚机制 团队成员 道德风险行为 多智能体强化学习 公共物品博弈模型 

分 类 号：C936[经济管理—管理学] F272.92