New AI model predicts epidemic spread with group-level immunity

A team of physicists including Yusheng Li and Meiling Xie has published a groundbreaking study in *Applied Mathematical Modelling* that leverages hypergraph theory and adaptive AI-driven interventions to model epidemic dynamics. Their work introduces an adaptive simplicial susceptible-infected-susceptible (s-SIS) model on d-uniform hypergraphs, where both individual infection states and group interaction patterns co-evolve in response to local infection pressure.

The model incorporates two novel intervention classes: risk-driven immunization—combining spontaneous isolation with targeted deactivation based on hyperedge infection pressure—and structural rewiring, which reconstructs group structures either randomly or via degree-preferential attachment. Using microscopic Markov chain approximations extended to higher-order interactions, the researchers derived analytical conditions for endemic and disease-free stationary states. Monte Carlo simulations confirmed that targeted interventions significantly outperform random strategies, suppressing epidemic prevalence and even inducing discontinuous phase transitions that drive systems toward disease-free equilibria.