IHBEM: Enhancing Influenza Forecasting Through an Integrated Platform for User-Generated Temporal Forecasts

In this project, a participatory modeling approach is taken to guide construction of a human judgment platform that generates temporal forecasts of the trajectory of an infectious agent. It is posited that to learn about the behavioral dynamics of experts three key features must be considered: (1) the factors that an expert uses to make decisions, (2) the accuracy by which any human and expert can predict an epidemic, and (3) how a set of forecasts can be combined to more accurately model the future to improve decision making. Work in the field of infectious disease modeling and human behavior typically concentrates on the public, often overlooking experts who make decisions that influence the general population. No work to date has explored constructing a human judgment forecasting platform that can collect temporal forecasts from individuals, methodology specific to combining human judgment temporal forecasts into an ensemble, and focusing on the characteristics of public health decision processes which impact downstream general population behaviors. Not only does this project advance the science of infectious disease forecasting, but it has the potential to benefit several populations who are at high risk for adverse outcomes due to influenza. The goals of this proposal are divided into three tasks. Task 1 involves recruitment of experts in the modeling of infectious disease, public health officials, and infectious diseases clinicians. From this population, cultural norms/needs are established and thought processes associated with infectious disease decision making are identified. In Task 2, a novel human judgment platform is constructed that experts and a lay audience can use to generate temporal forecasts of the trajectory of an infectious agent. This serves as a testbed to measure the performance of expert and lay temporal forecasts and compare human forecasts to computational model forecasts. Task 3 involves implementation and comparison of the performance of algorithms for combining human judgment forecasts; understanding the properties that are in common between human judgment and computational forecasts; and building a novel algorithm trained on traditional surveillance and augmented by human judgment temporal forecasts. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.