Reducing polarization in social networks with adversarial opinion perturbations

Abstract

The influence of perturbations on the evolution of opinions is a significant topic in complex systems and sociophysics. Previous studies have suggested that perturbations can alter opinion distributions, either promoting or inhibiting consensus formation. In this paper, we propose a model that integrates complex network topology and adversarial perturbations to simulate social opinion dynamics. Four main parameters are considered: social sensitivity, homophily, perturbation intensity, and network-connection probability. Our numerical simulations show that three primary patterns of collective opinion can emerge: consensus, radicalization, and polarization. Higher social sensitivity strengthens the radicalization and polarization states, while stronger homophily leads to more polarized opinions. Increased connection probability intensifies polarization by increasing network connectivity; however, adversarial perturbations reduce the prevalence of radicalization and polarization states, with stronger perturbations producing greater reductions. We employ microscopic evolutionary analysis to explain this pattern. These findings point to the important role of network structure and adversarial perturbations in reducing extreme states, offering theoretical insights for addressing societal polarization.