Inertia in Moral and Value Judgments of Large Language Models