Machine Learning Efficient Behavioral Interventions for Novel Epidemics

Policy makers that prepare responses to emergency situations need to have an understanding of the populations’ values, preferences and decision-making processes to implement effective interventions. Understanding the role of human behavior in epidemic outcomes has long been recognized as important, and is seen in the case of the COVID-19 pandemic. We will use a combination of data fusion techniques, large-scale agent-based simulation models, and inverse reinforcement learning to construct a framework that allows them to learn behaviors and responses to interventions so they can reshape responses and preferences  to  accomplish shared goals, including  saving lives and maintaining a sustainable economy. This GIDI seed grant will allow us to generate a framework that will answer these questions. In the pilot work, synthetic populations will be used to prepare a calibrated, analytics and simulation platform; in follow-up work, surveys and sources such as social media data will be used to fine-tune models.