Weighing the Trade-Offs of Explainable AI