Over at Philosophy, et cetera, Richard Chappell questions the common assumption that "one's metaethical views are more or less independent of one's first-order moral views." Chappell says that moral anti-realists act as if people really mattered, because people do matter to them. However, he says, that is not the same as believing that people matter in and of themselves. Can anti-realists believe that people matter simpliciter? Sure, they can act as if they do, but that is not the same as really believing it. If they don't really believe it, he says, then moral anti-realism may be morally suspect.
There is a brief but interesting discussion in the comments section of Chappell's blog. One good point which was raised is this: An anti-realist need not recognize a difference between acting as if people deserve respect and really believing that they do. In other words, anti-realists can be dispositionalists about belief. While that response to Chappell might satisfy some anti-realists, it might not satisfy non-cognitivists or expressivists, and anyway, it doesn't satisfy Chappell.
I think Chappell is probably wrong to reject this point too quickly. I also think he may be too quick to dismiss expressivism. In any case, I raised a couple of different issues in a comment. Here's what I wrote:
I wonder if your argument amounts to a rejection of consequentialism. If you agree that the moral anti-realist can act in all the same ways as the moral realist, but merely lacks some propositional attitude, then you can't be judging the moral worth of that attitude in terms of its consequences. You might say that the moral realist does act in some ways which are different: specifically, the moral realist avows moral realism. But does the mere avowal of moral realism produce a greater good (or reduce more suffering) than the avowal of moral anti-realism? If we are to be consequentialists, I think you need to find a bigger behavioral difference between moral realists and moral anti-realists to convince me that one is of more moral worth. If you were prepared to reject consequentialism, however, then your argument might be more persuasive. The moral anti-realist would do the same things as the moral realist, but would presumably be regarding people as means, and not just as ends. This would be a problem for a deontologist like Kant.
I think some moral anti-realists might be open to a limited reading of Kant, though. A moral anti-realist could believe that people are ends in themselves, and that there are facts in the universe which make it so. They might even say this is so in all possible universes. The argument could be that dignity is a necessary aspect of rational agency, which itself is a prerequisite for social contracts. Moral anti-realists can say it is impossible to conceive of rational agents without dignity. Thus, there is a fact of the matter which makes people ends in themselves.
I can see two ways the moral anti-realist can go about this. One is to claim that personhood is not a natural kind, and that there is no fact of the matter about whether or not an organism is (or should be treated as) a person. The other option is to claim that there is no fact of the matter about whether or not we should recognize the dignity of any particular person. So, the moral anti-realist can say that all persons have dignity, but the universe does not determine what is or is not a person. Or the moral anti-realist can say that there are objective persons, but there is no objective reason to respect the dignity of any particular person. In both cases, the moral anti-realist does genuinely believe in the dignity of persons. However, their moral attitudes do still admit of contingencies. But this does not seem so morally suspicious. It seems fair to suppose that personhood is not a natural kind. It also seems fair to suppose that there are circumstances in which a person loses their moral right to have their dignity respected. Neither of these views suggests that the moral anti-realist is too superficial in their attitudes or insincere in their behavior.