Selective Explanations: Leveraging Human Input to Align Explainable AI