Susan Wolf gave another interesting talk the other week, this time on the concept of 'moral obligation'. (We may take this as equivalent to notions of what is 'morally required', or what it would be 'morally wrong' not to do.) The most natural way to explicate the idea is in terms of what one has decisive moral reason to do, but Wolf suggested that this doesn't work.
The problem is that we often have decisive moral reason to do things which (seemingly) aren't morally obligatory. Wolf appealed to the example of not driving an SUV. The environmental and safety disadvantages of SUVs count against them, and for typical urban usage there really aren't sufficient favourable reasons to counterbalance these and make driving an SUV around town a reasonable thing to do. Sure, it's not the end of the world, but the moral reasons here do count conclusively against driving an SUV. It's not something a perfectly reasonable agent would do. Now, despite this, we do not typically think people are obligated or required to drive more sensible cars. While their decision may be morally imperfect, it is not "immoral" in the strong sense (which we may align with blameworthiness and social censure). So decisive moral reasons are insufficient to establish moral obligation.
I wonder whether decisive moral reasons are at least necessary for obligations to be incurred. Perhaps this must be so, in order to leave room for supererogatory actions. We can say that an action is supererogatory when the strong moral reasons that count in its favour are nevertheless not decisive (perhaps due to strong prudential disadvantages).
This leaves us with a tripartite moral structure. An agent might reasonably fail to perform supererogatory actions, though it would be very nice if they did manage them. Then there's the important middle layer where one has decisive moral reasons, and so is at least unreasonable (in the weak sense of "less than ideally reasonable") if one fails to act accordingly. And then we have the base level of moral obligation, or what is required to meet minimal standards of moral decency. This reminds me of the "ethical minimalism" Paul Studtmann argued for once back at Canterbury. Though I got the impression that Wolf considers the second level to be more important (and appropriate to aspire to) than the minimalist base.
But I digress. Returning to Wolf's talk: she pointed out that we need to set aside a small subset of the morally desirable actions as 'obligatory' for pragmatic reasons. There are too many morally desirable actions, and we can't expect everyone to satisfy them all. That would make morality too demanding. So it is useful for society to be able to point to a subset of the most important actions and say, "you must at least do those!" It is the binding force of this 'must' which distinguishes moral obligation from the weaker sense of moral desirability in which you ought not to drive an SUV.
The crucial question now arises: how are we to draw this distinction? What makes some morally desirable actions obligatory, and not others? One might initially think to appeal to the 'weightiness' of the moral reasons. (The SUV case seems non-obligatory precisely because it is relatively trivial. The reasons are decisive, but decisively small.) But that won't do, because we can have trivial moral obligations, such as the obligation not to steal a paperclip.
Wolf proposed a modified "social command theory" of obligation, such that X is morally obligatory only if X is commanded by society (and backed by adequate moral reasons). But that strikes me as unacceptably arbitrary, despite the parenthetical constraint. Surely the facts about moral obligation must be determined solely by morally relevant facts, i.e. facts about welfare, not anyone's arbitrary "commands". It also has the odd consequence that we can change what's truly obligatory, simply by influencing opinions or expectations, and hence altering "what society commands".
Besides, it isn't clear that anything else in Wolf's argument leads to this particular theory of obligation. In light of her pragmatic motivations, all she needs is some way or other to draw a distinction. Arbitrariness doesn't matter for her purposes, because she doesn't believe there's any principled basis upon which to draw the distinction in any case! So we might just as well adopt the Coin Theory of Obligation: for any morally desirable act-type A, flip a coin. If the coin lands heads, then A is morally obligatory. Otherwise it is not. (It's no more arbitrary than appealing to societal "commands", after all!)
In fact, given the pragmatic motivations, Wolf really should be led to an indirect utilitarian theory of obligation. Since our aim is to draw a distinction which will help promote more moral behaviour in practice, the obvious basis for this distinction is to identify that class of actions which, if recognized as 'moral requirements', will have the morally best consequences. There is certainly some fact of the matter about which such classes would have the best results, and so we have a principled basis for determining (in the metaphysical sense; whether we can know these facts is another question!) which actions are morally obligatory.