E.G.

Over time, metaethicists have shifted from

Definition 1: Moral realism is the view that moral facts are mind-independent

to 

Definition 2: Moral realism is the view that moral facts are stance-independent.

I believe that this shift was motivated by recognition that moral facts clearly aren't entirely mind-independent. For example, the painfulness of torture is part of what makes torture wrong. Realists are not and have never been committed to deny that moral facts depend on facts about pain. Yet pain is mental. So, Definition 1 was always wrong.

What Definition 1 got right (and perhaps also what made it seem good enough for government work) is that Definition 1 correctly implies that realism is at least in part a mental independence thesis. That is, realism says that there is some class of mental phenomena such that moral facts are independent of the items in that class. (That might not be all that realism says, but it is one of the things that realism says.) The problem was that Definition 1 defines that class far too inclusively.

So, Definition 2, I believe, was brought in to correct the error of Definition 1 by naming the relevant class of mental phenomena. That is, stances are supposed to comprise the class of mental phenomena that (according to realism) moral facts are independent of.

But what are stances? 

I may be mistaken about this but I believe that Russ Shafer-Landau is the person who caused metaethicists to shift from Definition 1 to Definition 2. So, if we want to understand the notion of stance, we might start by looking at what Shafer-Landau and his coauthors Terence Cuneo and John Bengson say about stances in their recent agenda-setting book, The Moral Universe.

B, C, and S-L say that one of the theses of realism is

Stance-­Independence: A range of moral truths and facts are stance-independent.

Clarifying this, they say:

The Stance-­Independence thesis is intended to convey the idea that a range of moral truths or facts are not of our own devising; unlike the norms, say, of badminton and bowling, they do not depend in any way on our (even idealized) say-­so.

And they say we can

think of a truth or fact as being stance-­independent just if there is a full metaphysical story about it that is stance-­free—­where a stance is an attitudinal state or activity (or pattern thereof), whether actual or counterfactual, borne by an agent. So understood, stances include states such as approval or disapproval, and explain phenomena such as speech acts, intentional behaviors, and social arrangements.

I doubt that 'stance' should be understood in the above way if we are going to hold on to Definition 2.

Suppose you disapprove of my using your lawnmower, but I do it anyway. Here one might say that it is morally wrong for me to use your lawnmower because you disapprove of my doing so. In saying this, one does not commit oneself to anti-realism. But saying this would come with a commitment to anti-realism if disapproval is a type of stance and if Definition 2 is correct. So, we should either not say that disapproval is a type of stance or shift away from Definition 2. My hunch is that we should keep Definition 2, at least for now, and work on refining the notion of stance.

It seems to me possible that we should understand realism as a vague cluster of separable independence theses. Perhaps we should start with something like this: According to realism, if p is a true moral proposition, then the fact that p is independent of any and all beliefs about whether p. Some independence thesis like this is, perhaps, at the center of the cluster. And then perhaps there are other independence theses that are clearly in the cluster, while other independence theses are closer to the edge, and still others clearly beyond the edge.