When to Make Exceptions: Exploring Language Models as Accounts of Human Moral Judgment