arXiv:2610.41943v1 Announce Type: new Abstract: The COVID-19 pandemic has exhibited multiple waves of infections, reflecting feedback between disease transmission and adaptive population behavior. Accurately predicting and understanding these dynamics remains a central challenge for epidemic modeling. In this paper, we develop a forecasting framework based on a Susceptible-Exposed-Infectious-Recovered (SEIR) model coupled with an evolutionary model of behavioral adaptation, in which population compliance with protective measures evolves in response to perceived infection risk and associated costs. We formulate forecasting as the estimation of behavioral incentive parameters that govern the coupled behavior-disease dynamics, using hospitalization time series data together with standard assumptions on disease-related parameters. Using the inferred parameters, we generate rolling monthly forecasts of infection levels and evaluate the approach on both synthetic data and hospitalization time series from multiple U.S. states (Texas, California, Colorado, and Arizona). The proposed framework provides a principled way to incorporate behavior-disease feedback into epidemic forecasting while yielding interpretable estimates of population-level incentives.