Modeling and Simulation of Evolutionary Learning Processes of Organizations

从基于智能体的建模角度出发，提出了进化学习模型。特别采用了两个方面的建模方法：超博弈模型和分布模型。在超博弈模型中，利用树人网络型动态超博弈，系统地考察了所提出的学习过程模型在更一般的收益设置下的有效性。在分布式模型中，提出了一种基于单环学习和双环学习层次的新型分布式模型，该模型以组织学习为基础。该模型是对前模型的扩展，是对超博弈情境的推广。在该模型的双环学习过程中，智能体的情境感知学习在所有逻辑可识别的情境中都成功完成。将提出的组织学习模型应用于消费者群体中偏好函数的演化和学习过程，并利用遗传算法进行仿真分析。该模型为分析消费者群体的适应过程提供了一个框架。最显著的结果之一是可以获得潜在优秀消费者的特征。研究成果在北美计算社会与组织科学学会、亚太社会系统科学基于agent方法学会、日本仪器与控制系统学会等国内外学术会议上发表。

Formulations of evolutionary learning models have been done from the viewpoint of agent-based modeling. In particular two aspects for modeling have been employed : hypergame type model and distributive model.In the hypergame type model, using tree-person network-type dynamic hypergame, the effectiveness of the proposed learning process model for more general settings of payoffs were systematically investigated.In the distributive model, a new type of distributive models were proposed, which are based on the levels of single-loop learning and double-loop learning as the basic organizational learning. The proposed model is substantially extended from the former model and a generalization of hypergame situation. The learning of agents' situation perceptions in the double-loop learning process of this model was successfully performed in all of the logically identifiable situations.As an application of the proposed learning model of organizational learning, the evolutionary and learning process of preference functions in a consumer population was modeled, and analyzed in simulation using genetic algorithm. The model provides a framework of analyzing the adaptation process of the consumer population. As one of the most remarkable results, the characteristics of potentially excellent consumers could be obtained.The research results were presented at some internal and domestic academic conferences such as North America Association for Computational Social and Organizational Science, Pacific Asia Association for Agent-based Approach in Social Systems Science, Japan Society for Instrument and Control Systems and so on.