This document discusses combining human learning and machine learning to improve recommendation solutions. It proposes starting with a heuristic human approach, then adding machine learning methods like scoring and feature selection over time. Multiple algorithms, both human learning and machine learning, would be tested alongside each other in a "King of the Hill" approach. The goal is to develop the best performing solution by leveraging both human knowledge and machine learning capabilities. Feedback would be used to refine the models on an ongoing basis.