Optimizing Interventions for Agent-Based Infectious Disease Simulations

A team of researchers has developed a novel AI system called ADIOS (Agent-based Infectious Disease Intervention Optimization System) to tackle one of public health's most complex challenges: finding the least disruptive yet effective measures to control pandemics. The core problem is that the search space for Non-Pharmaceutical Interventions (NPIs)—such as targeted school closures, workplace restrictions, or isolation rules—is vast or even infinite in detailed agent-based simulations. ADIOS structures this chaos using a domain-specific language and employs Grammar-Guided Genetic Programming (GGGP), an AI optimization technique, to efficiently evolve and test intervention strategies.

The system was demonstrated using the realistic German Epidemic Micro-Simulation System (GEMS). By defining constraints to prevent nonsensical policies, ADIOS can automatically generate and evaluate combinations of interventions that target individuals based on multiple attributes (like age or profession) and affect hierarchical group structures (like families or schools). This moves beyond manual, scenario-based testing, offering a data-driven method to support policymakers. The goal is to identify optimal strategies that balance disease suppression with minimizing economic and social costs, providing a crucial tool for preparedness when pharmaceutical options are not yet available.