EpiWorld integrates language models with epidemic simulations for better policies
EpiWorld: Grounding LLM Policy Agents in Epidemiological World Models
Computation and Language
Summary
Making decisions about how to control an outbreak is complicated because you need to predict how illnesses will spread and how different actions will help. The authors created EpiWorld, a system that combines a type of AI that understands and generates language with a model that simulates how diseases evolve. This system lets the AI test different strategies and learn from the outcomes to improve decisions over time. They tested EpiWorld using data from COVID-19 and the flu, finding it could reduce hospitalizations more effectively than some other decision methods.
What this means in practice
- •For public health policymakers: Use EpiWorld to evaluate and refine epidemic intervention plans by simulating disease spread and policy outcomes before implementation.
- •For healthcare operations managers: Improve hospital resource planning by forecasting hospitalization trends under different intervention scenarios simulated by EpiWorld.
Authors
Zeeshan Memon, Yiqi Su, Kai Shu, Naren Ramakrishnan, Liang Zhao
Abstract
Epidemic intervention policies are textual artefacts that human decision-makers interpret, justify, and revise through natural language, making large language models a natural candidate for epidemic policy reasoning. A naive LLM, however, lacks the epidemic dynamics needed to project intervention consequences, the quantitative surveillance signals required to assess severity, and the institutional constraints that define admissible actions. We present EpiWorld, a closed-loop framework that grounds an LLM policy actor in a learned action-conditioned epidemiological world model and a tiered skill library of public-health protocols, surveillance tools, and adaptive lessons accumulated through after-action analysis. Given a candidate intervention, the world model predicts regional epidemic evolution and enables fast counterfactual rollouts that provide feedback for policy selection and refinement. Outcomes of simulated futures are distilled into reusable lessons while protocol constraints remain fixed, allowing the decision process to improve without sacrificing interpretability or controllability. We evaluate both the world model and the end-to-end framework on retrospective COVID-19 and Influenza datasets: the world model achieves the best out-of-distribution Peak-MAE among all forecasting baselines, and the closed-loop framework reduces cumulative hospitalisation by up to 59% across datasets and by an average of ~16% across six LLM backbones, outperforming reinforcement-learning and optimal-control policy baselines.