Opinion formation and bi-polarization with biased assimilation and homophily

An agent-based model incorporating biased assimilation is proposed in this paper to investigate opinion dynamics over a connected social network. The opinion of each agent is represented by a sequence of arguments, and it evolves through the interactions between agents. The probability that one agent chooses another to communicate depends on the similarity of their opinions. During every interaction, interacting agents exchange the argument randomly selected from the corresponding arguments sequence. Theoretical analysis reveals that this model results in consensus on either extreme positive opinion or extreme negative opinion, or generates bi-polarization. Numerical simulations are carried out to investigate the dynamics of the model over different networks. Results are obtained in terms of the effect of homophily, biased assimilation and network topology on opinion formation.