Graphical Evolutionary Game Theoretic Modeling of Strategy Evolution Over Heterogeneous Networks

The rapid development of the Internet and networking technologies greatly facilitates the interactions among heterogeneous agents, while it also causes problems, such as the breakout of rumors and online bullying. Thus, it is critical to study the strategy evolution process in complex networks, that is, how heterogeneous agents interact with each other, update their opinions, and make decisions. In the literature, there have been numerous works on the modeling and analysis of decision making and opinion dynamics in social networks. However, most works assume that agents are homogeneous or only consider one single attribute in agent heterogeneity. In complex social networks, agents differ in many attributes and they constantly influence each other's decisions. How strategy evolves in complex networks with heterogeneous agents remains unknown. In this work, we consider three different attributes of agents: influence, susceptibility, and interest. We use graphical evolutionary game theory to theoretically analyze the impact of different attributes on the strategy evolution process and the evolutionary stable states (ESS). Both theoretical analysis and simulation results show that the influence attribute alone cannot change the ESS of the strategy evolution, while super agents who are both influential and stubborn have the largest impact on the ESS. Furthermore, real data validation shows that our proposed model can effectively model the information diffusion process in online social networks. This study is critical to the better understanding of agents' decision making process, and provides important guidelines on the management of social networks.