【作者】 施永仁；

【导师】 高亮；

【作者基本信息】 华中科技大学 ， 工业工程， 2007， 硕士

【摘要】 社会经济系统、生态系统、制造系统等复杂巨系统由于其内部高度的非线性、关联性和分散性,使得传统的数理方法无力解决这类系统宏观上涌现的整体结构与行为。复杂性科学旨在从整体上把握系统的结构与行为,抛弃了还原论思想,自下而上地研究其集体涌现行为。20世纪90年代,圣菲研究所提出的复杂适应系统理论成为当代复杂性科学的理论基础,其方法论——基于Agent的建模方法——大大方便了社会科学工作者对社会系统进行实验与分析,有助于理解社会经济系统的运行。本文首先对复杂性科学和复杂系统理论进行了简要回顾,概述了其主要理论分支与研究方法,给出了一个研究大背景。并着重介绍了复杂适应系统(Complex Adaptive Systems, CAS)理论及基于Agent的建模方法(Agent Based Modeling, ABM),分析了它们在社会经济系统的应用前景与研究思路。针对社会经济系统的一些现象,本文将CAS理论应用于社会经济系统,并分别分别给出了三个应用实例。第一个实例是人工经济模型,在虚拟世界中构造一个由适应性Agent组成的经济系统,Agent之间进行着交易与学习,交易网络和资源传播网络以“无标度”的拓扑形式涌现出来,在市场中少数交易者控制着大部分的交易。第二个实例解释了自私的企业为何会自发的涌现出对社会的责任,通过对企业建立广义Lotka-Volterra模型构建了系统的多Agent模型,发现自私的企业通过相互之间的竞争而逐渐地开始采纳“利他的”企业社会责任这一行业规范。第三个实例以社会系统中的传播现象为对象,分别研究了小世界上的简单传播,复杂传播和多元文化传播。研究结果表明,不同类型的传播在小世界拓扑结构的社会网络上表现出了完全不同的行为,复杂传播会因加边概率的增加而减小传播成功的几率,但传播速度会显著地增加;而多元文化传播的研究则显示全局上文化的沟通不但不会促进不同文化集团之间的融合,反而会加剧他们间的对立,这一结果能加深对全球化和文化间的冲突的理解。本文最后还对多Agent仿真平台进行了介绍,本文从理论到方法到应用全方位的对复杂适应系统及其方法论——基于Agent的建模方法——进行了论述。复杂适应系统理论提供了一个全新的视角,其系统的层次性,非线性、涌现性以及分化性等复杂系统特征在一定程度上弥补了传统社会科学研究方法的不足。本文从理论到方法到应用全面的论述了复杂适应系统与基于Agent的建模方法在社会经济系统中的实践,为更深层次的工作起到铺垫作用。更多还原

【Abstract】 Due to the inherent nonlinearity, coupling and distributedness of Complex Social Systems, the traditional mathematical methods are invalid to handle the emergent characteristics and macro behavior of the complex systems. Complexity Science aims to grasp the structure and behavior of large scale systems. It abandoned the traditional reductionism, which asserts that the nature of complex things is reduced to the nature of sums of simpler or more fundamental things. As a foundation of Complexity Science, Complex Adaptive Systems provides a metaphor to the operation of complex system and its methodology—the agent based modeling—facilitates the research and analysis of large-scale systems.This paper traces back the history of complex system and complexity science to summarize the primary schools of system science and its research methods. Chapter 2 emphasizes the theory of Complex Adaptive Systems and the Agent Based Modeling methodology.Based on the research background and the complex phenomena of social-economical system, three computer simulations were built to validate the application. First model is the Artificial Economy Model, which is a virtual world composed of numerous adaptive agents. The Scale free Transaction Network and propagation network emerge out of the local interations of adaptive agents. Second model explained how a bundle of selfish firms compete with each other to emerge an altruism Corporate Social Responsibility(CSR). This model is built on the generalized Lotka-Volterra model which is suitable for multi agent competing system. The third model is proposed to investigate the contagion process on the“small-world”social networks. Simple contagion, complex contagion and multi-dimensional culture disseminations are well explored. The results showed that more links or global interations introduced in the networked systems, more rigorous these culture groups opposed to each other! Globalization can not accelerate the understanding of different cultures but even to increase their gap!The multi agent simulation platforms are introduced at the end of the thesis. The complex system characteristics, such a hierarchy, nonlinearity, emergence and differentiation and so on, got by its model offset the shortage of traditional economics research to a certain extent. This study is of active implication for the advanced application of the theory in China and could hopefully provide some experience technical support for future application of multi agent simulation technology.更多还原