Multi-population opinion dynamics model

We propose a novel multi-population opinion dynamics model that extends the bounded confidence model to analyze how individual and group interactions influence the emergence of consensus, polarization, or fragmentation. Traditional models often compromise between accuracy and scalability by either overlooking agents’ similarities or incorporating detailed agent-level connections at the expense of broader applicability. Our model addresses this challenge by leveraging group affiliations to define agent similarities in a scalable manner. The Wasserstein distance is used to quantify the closeness between sub-populations. The dual nature of the proposed method enables the examination of micro-level (individual-level) interactions and macro-level (group-level) dynamics, offering insights into the reciprocal influence between individual behaviors and group dynamics. We rigorously prove the well-posedness of our models and employ simulations to mimic complex social phenomena, demonstrating the framework’s ability to capture the social mechanisms driving public opinion formation.