On The Role of Community Structure in Evolution of Opinion Formation: A New Bounded Confidence Opinion Dynamics

The process of opinion formation and evolution is an important part of the study of opinion dynamics. However, the existing opinion dynamics ignore the relationship between opinion evolution and community structure. To address the limitation, we build a novel bounded confidence opinion dynamics model (called LPA-HK for simplicity). Based on the Hegselmann-Krause (HK) model, the label propagation algorithm (LPA) is introduced in the LPA-HK model to study the relationship between opinion evolution and community partition from the micro-perspective of the evolutionary process. Each agent has an opinion as well as a label indicating the agent’s community. The iterations of agents’ states involve two stages: label and opinion updates. The labels and opinions of agents evolve dynamically until the group reaches a stable state. We perform simulations to examine the effect of the model parameters on opinion evolution and community partitions. We show that group reaches a consensus as the confidence level increases to a critical value and the partitions of high quality are obtained when the confidence level is greater than the critical value. We apply the LPA-HK model to three kinds of real networks to solve the community detection problem. The effectiveness of the LPA-HK model is verified by comparing the results with two existing community detection results based on modularity optimization algorithms.