Rule-Based Moral Principles for Explaining Uncertainty in Natural Language Generation