Learning, Game Play, and Convergence of Behavior in Evolving Social Networks
MetadataShow full item record
I study information dissemination and opinion formation in a framework of evolving social networks. Individuals take weighted averages repeatedly to update their opinions. They also update their assessments on others' opinions, represented by an influence weight matrix. It is proven that both opinions and the influence weights are convergent. In the steady state, consensus is reached where all individuals hold the same opinion. Convergence occurs with an extended model as well, which indicates the tremendous influential power possessed by a minority group. Then I impose a dual network structure, where individuals not only collect information, but also use the information to play a coordination game with a selected group of opponents that one is connected with. All individuals update their strategies based on a naive learning process within a separate influence network in which information is disseminated. The selection of opponents also gets updated over time. I calculate the critical values of costs associated with connections for different network structures and strategies to occur in the steady state. Finally, I investigate the outcomes of social learning under various exogenous network structures. Individuals use an algorithm that takes into account both proximity of opinions and impact of neighbors. Results also show consensus, with convergence speed correlated with the network structure. In addition, an endogenous network formation in two stages that utilizes network and distance between agents' opinions is proposed. The resulting networks show power-law patterns in degree distribution.
- Doctoral Dissertations