World models can help epidemiologists reason about hidden disease burden, account for policy-dependent surveillance bias, and simulate counterfactual intervention outcomes—capabilities essential for evidence-based epidemic control.
This paper proposes using world models—AI systems that learn to simulate how systems evolve over time—for epidemiology. The authors argue that epidemic decision-making is uniquely suited to world models because disease spread involves hidden states, noisy observations that depend on policy choices, and interventions that trigger behavioral responses.