A Survey on Explainability: Why Should We Believe the Accuracy of A Model?