Explainable AI through the Learning of Arguments