Utrecht University, Department of Sociology/ICS, Utrecht, The Netherlands.
Utrecht University, Institute for Theoretical Physics, Utrecht, The Netherlands.
Sci Rep. 2017 Dec 5;7(1):17016. doi: 10.1038/s41598-017-16982-2.
Network structure can have an important effect on the behavior of players in an iterated 2 × 2 game. We study the effect of network structure on global and local behavior in asymmetric coordination games using best response dynamics. We find that global behavior is highly dependent on network topology. Random (Erdös-Rényi) networks mostly converge to homogeneous behavior, but the higher the clustering in the network the more heterogeneous the behavior becomes. Behavior within the communities of the network is almost exclusively homogeneous. The findings suggest that clustering of networks facilitates self-organization of uniform behavior within clusters, but heterogeneous behavior between clusters. At the local level we find that some nodes are more important in determining the equilibrium behavior than other nodes. Degree centrality is for most networks the main predictor for the behavior and nodes with an even degree have an advantage over nodes with an uneven degree in dictating the behavior. We conclude that the behavior is difficult to predict for (Erdös-Rényi) networks and that the network imposes the behavior as a function of clustering and degree heterogeneity in other networks.
网络结构对重复 2×2 博弈中参与者的行为有重要影响。我们使用最佳响应动力学研究了网络结构对非对称协调博弈中全局和局部行为的影响。我们发现，全局行为高度依赖于网络拓扑。随机（Erdős-Rényi）网络主要趋向于同质行为，但网络聚类度越高，行为的异质性就越强。网络社区内的行为几乎完全是同质的。研究结果表明，网络聚类有助于在聚类内部自组织一致的行为，但在聚类之间则产生异质的行为。在局部水平上，我们发现有些节点比其他节点在确定均衡行为方面更为重要。在大多数网络中，度中心性是预测行为的主要指标，具有偶数度的节点比具有奇数度的节点在决定行为方面具有优势。我们的结论是，（Erdős-Rényi）网络的行为难以预测，而网络则将行为作为聚类和度异质性的函数强加给其他网络。
