| The evolution of the public opinions and the network structure are prominent in both real-world populations and virtual communities. That is to say, in a social network, one is prone to be influenced by her neighbors so as to align her own opinion with the neighbors’ and this illustrates the impact of network on opinions. In turn, the influence of the opinions to the network structure cannot be neglected either. It is common that people with similar opinions tend to link together while dissimilar persons tend to be separated. The research of co-evolution of opinion and network is therefore deserved and this research direction is potentially meaningful for social marketing in business.The co-evolutionary dynamics of opinion and network is a typical problem of complex adaptive systems; therefore the agent-based modeling approach is adopted in this thesis. Firstly, a multidimensional discrete vector is used to represent the opinion in the model, in order to cope with the multi-dimensionality of opinions in the real-world. Secondly, opinion exchange takes place between network neighbors whose opinion similarity is higher than a preset threshold, to mimic the real-world situation of bounded confidence in communications. At last, we define the rule for agent migration in accordance with the mechanisms of "homophilous linking" and social cohesion. With this agent-based model, computational experiments are conducted and the results are summarized as follows.First, from the aspect of opinions, community opinion will converge to some degree. Stronger the social influence is, more easily the consensus can be reached. When the agents in the model communicate with each other along with neighborhood adjustment, the convergence of opinions would be enhanced. This indicates that the bi-directional influence is significant for the adjustment of neighborhood and the contagion of opinions. Meanwhile, it is observed in the simulations that the co-evolutionary dynamics would result in the formation of opinion clusters in which the agents with identical opinions are densely inter-linked. In particular, it is found that the randomness in selecting an interactive partner would increase the unevenness of the sizes of opinion clusters.Second, from the aspect of network structure, in the process of co-evolutionary dynamics, the network may evolve to form a small-world network which is composed of multiple inter-linked opinion clusters. The analysis shows that the mechanism of homophilious linking has a prominent influence on the formation of such small-world network, while the structural cohesion is the secondly important factor. In summary, in this thesis the coevolutionary dynamics of multidimensional opinions and social network is studied, resulting a few interesting findings. In particular, the coevolution would head for a small-world network of opinion clusters which are weakly interconnected, owing to the microscopic mechanisms of homophilious linking and social contagion. |
