个体异质性视角下双群体连接与行为共选博弈学习动力学研究

With the progress of science and technology, agents can participate in interaction in a much deeper level and on a broader scale, and the differences between agents tend to play an important role. As a result, the interaction between agents choosing both actions and links becomes more complicated. The literature of learning dynamics in games with agents choosing both actions and links usually assumes that all agent are homogenous, and cannot analyze the interaction between agents in the setting of heterogeneity. To solve this problem, this project introduces the assumptions of heterogeneous action sets, bilateral interaction between two groups of agents, heterogeneous preferences over actions and dual action and heterogeneous bounded rational learning rules. Firstly, the project establishes randomly perturbed learning dynamics in two-population games where agents choose both actions and links by extending the network game theory. Then, by applying the theory of learning in games, the project identifies stochastically stable equilibria, evaluates the rate of convergence to stochastically stable equilibria and explores the impact of heterogeneity on agents' long-run behavior. The project can offer novel framework of dynamical analysis for the practices in economics and management related to interactions between heterogeneous agents and can help government and academic researchers understand various phenomena.

随着科技的飞速发展，行为主体可以在更深层次和更广范围参与互动，主体间的差异造成的矛盾日益凸显。连接与行为共选博弈学习理论是分析上述应用实践的有力工具，然而现有理论模型一般假定“所有个体为同质的”，在分析有差异个体间的复杂互动上遇到瓶颈。据此，本项目拟引入“个体异质性”视角，围绕连接与行为共选博弈学习动力学模型主要组成要素，提出四个新的假定：异质行为策略集、双边互动、异质行为策略偏好与兼容策略、异质有限理性策略调整规则。首先，借鉴网络博弈理论最新研究进展，针对上述四个假定，分别构建双群体连接与行为共选博弈学习动力学模型。然后，运用博弈学习理论的方法与技术，求解随机稳定均衡，估算收敛到随机稳定均衡的速度，考察四个方面的异质性对个体长期行为的影响。本项目理论模型将为经济管理领域中异质个体互动的应用实践提供全新的动态分析框架，提升学界、产业界等不同主体对诸多经济与社会现象背后客观规律的认识与把握。

随着信息通讯技术的发展，人们日渐清晰地认识到，各种经济管理活动均发生在一个相互连接的世界中。连接与行为共选协调博弈是分析上述应用实践的有力工具。通过引入连接、信息、潜在博弈等异质性，本项目主要取得如下成果。. 首先，本项目区分主动连接和被动连接，研究被动连接如何在连接与行为共选协调博弈中造成效率损失。由于被动连接的影响，在Nash均衡和随机稳定均衡中，不同个体可能采用不同动作，并且，采用不同动作的个体也可能发生互动。研究结果《Lock-in through passive connections》已发表于理论经济学顶级期刊Journal of Economic Theory。然后，本项目引入个体信息观察异质性，假定：全部个体分布在一个任意给定网络，两两匹配为对，参与2人协调博弈。个体通过模仿调整策略。本项目辨明随机稳定均衡为Pareto最优策略组的充分条件。研究结果《Matching, imitation, and coordination in networks》已发表于具有重要影响的博弈论期刊Dynamic Games and Applications。随后，本项目引入潜在协调博弈的异质性，将规模效应纳入连接与行为共选协调博弈。通过刻画从众效应，本项目辨明了，行为策略间互补性如何促使最有效的行为策略扩散。相应研究成果《Bandwagon effects and constrained network formation》正在博弈论顶级期刊Games and Economic Behavior修订（R&R）。最后，本项目引入个体能否建立连接的差异，给出一个基于双边互动的连接与行为共选协调博弈学习动力学模型，刻画了全部随机稳定均衡。研究结果《Social coordination and network formation in bipartite networks》已发表于经济学主流期刊The B.E. Journal of Theoretical Economics。另外，本项目所资助的研究成果发表于International Journal of Game Theory、Theory and Decision、《运筹与管理》等国内外重要中英文经济学和管理学期刊。本项目理论研究成果将为经济管理领域中个体间复杂互动的应用实践提供了全新的动态分析框架。.