Enhancing predictive accuracy in social contagion dynamics via directed hypergraph structures
Juyi Li, Xiaoqun Wu, Jinhu Lü, Ling Lei
Abstract
Abstract Evidence from both theoretical and empirical studies suggests that higher-order networks have emerged as powerful tools for modeling social contagions, such as opinion formation. In this article, we develop a model of social contagion on directed hypergraphs by considering the heterogeneity of individuals and environments in terms of reinforcing contagion effects. By distinguishing the directedness between nodes and hyperedges, we find that the bistable interval of the discontinuous phase transition decreases as the directedness strength decreases. Additionally, directed hypergraphs tend to generate bistable intervals when nodes with a large hyperdegree are more likely to adopt a specific opinion, as evidenced by simulations of directionality assignments for three sets of real networks. These findings provide two approaches to enhance the accuracy of predicting social contagion dynamics: one is to increase the stubbornness of all individuals, and the other is to prioritize increasing the stubbornness of highly influential individuals.