基于Agent仿真的团队重组与集体问题解决关系研究

Study on the effects of Groups Recombination on Collective Problem Solving: An Agent-Based Simulation Approach

摘要:

个体由于认知能力的限制,通常集体完成复杂任务,表现出群体智能的现象。而如何提高群体智能一直是重要的研究问题。基于NK模型建立多主体仿真实验,本文探究隔离团队重组对集体问题解决能力的影响。仿真结果表明,团队重组频数与集体表现呈倒U关系,即适当的团队重组可以提高集体问题解决的能力,同时与半隔离团队对比发现,团队重组可以媲美半隔离团队的功能,并且在集体表现上具有更强的稳定性,因此团队重组具有更强的操作性。此外,权变理论分析发现,不存在团队重组时,群体应当保持较大的团队规模,较小的人员流动,较小的个体学习速率；存在团队重组时,群体应该保持较小的团队规模,较小的人员流动率,适当的个体学习速率。此外,在复杂问题面前,团队重组对提高群体解决问题能力的作用更为显著。

Abstract:

With the help of an Agent-based simulation based on NK model, we study the impacts of isolated-groups recombination on collective performance. The results of simulation experiments suggest that there is always an inverted-U relationship between recombination frequency and collective performance. At the same, compared with semi-isolated groups, groups recombination can achieve the same performance and the stability of the performance is more strong, therefore groups recombination is more easy to operate. Besides, when lack of recombination, crowds should keep larger group size, less turnover, lower learning rate; Otherwise, crowds should keep smaller group size, less turnover, and moderate learning rate. At last, it is easy for crowds with isolated-groups recombination to solve more complicated problems.