We develop a causal acceptance model for organ offers that learns from directional refusal reasons. Our method, ClexNet, leverages these explanations to guide learning and improve generalization beyond observed allocation policies.