On Aligning Prediction Models with Clinical Experiential Learning: A Prostate Cancer Case Study