Contextualized Policy Recovery: Modeling and Interpreting Medical Decisions with Adaptive Imitation Learning

The Contextualized Policy Recovery (CPR) framework offers a new approach to modeling medical decisions by adapting to patient-specific contexts. Unlike traditional methods that trade off accuracy for interpretability, CPR uses multi-task learning to generate context-specific decision models. This method improves the understanding of complex medical decisions by incorporating contextual information, leading to better prediction of antibiotic prescriptions in ICUs and MRI orders for Alzheimer's patients. CPR bridges the gap between black-box and interpretable models, offering both high performance and clarity in medical policy learning.