"Why did you do that?": Explaining black box models with Inductive Synthesis