Abstraction, Validation, and Generalization for Explainable Artificial Intelligence