AI Framework 'Neetyabhas' Optimizes Uncertain COVID-19 Policies with Agent Simulations

A team of Indian researchers led by Janani Venugopalan has introduced Neetyabhas, a novel AI framework for uncertainty-aware public policy optimization. Unlike traditional models that assume perfect infection tracking and flawless policy execution, Neetyabhas integrates real-world uncertainties in epidemic measurement (infections/hospitalizations) and policy implementation. The framework uses hierarchical reinforcement learning agents—specifically deep Q-networks alongside uncertainty-aware policy gradient variants (DDPG and TD3)—to simulate 1,000 individuals making dynamic choices about mask-wearing, vaccination, and shopping. Policymakers in the model deploy interventions like lockdowns and mandates based on noisy health and economic observations.

The simulations demonstrated that the dynamic control approach effectively managed epidemic progression. Masking and vaccinations were standout measures, significantly reducing both the outbreak's peak height and duration. By embedding individual behaviors and imperfect data into the policy framework, Neetyabhas overcomes limitations of prior research that overlooked human agency and real-world errors. The findings underscore that accounting for individual choices and incomplete information is critical for designing effective interventions during complex pandemics, positioning masks and vaccines as pivotal tools even under uncertainty.