I am somewhat puzzled by a recent post by Barry Arrington at "Uncommon Descent," which I came across through John Pieret's Thoughts in a Haystack. It lays out something like this argument:
(1) Atheistic naturalism is true. (assumption)
(2) One can’t infer an "ought" from an "is." (assumption)
(3) All that is is the natural world, and the natural world is all there is. (from 1)
(4) There is nothing in the natural world from which we can infer an "ought." (from 2 and 3)
(5) For any action, there is nothing from which one can infer that one ought to refrain from performing that action. (from 4 and 3)
(6) For any action, it is permissible if and only if it’s not the case that one ought to refrain from performing that action. (assumption)
(7) For any action, it’s permissible to perform that action. (from 5 and 6)
No one should actually accept (2) in an unqualified form; the original Humean reasons for the maxim don't justify an unqualified form, and neither, I think, do open question arguments; whereas there is excellent reason to think that an unqualified version of it should be rejected. But one can reasonably set this aside; (2) is widely accepted, and there are probably quite a few people who would indeed take it in an unrestricted way, not having thought the matter through. I still have two puzzles about this argument.
Puzzle 1: I can see no plausible reading of (1), or even (3), that makes it impossible for "the natural world" to include some fairly straightforward oughts, even if they are taken as primitives. For instance, most atheistic naturalists, I take it, would accept the view that there are obvious conditional oughts identifying optimal strategies for dealing with various problems: If you want to survive, you ought to do this; you ought to do that in order to minimize the possibility of harm to yourself and your kin. If each of these is taken as an 'is', then the argument really amounts to a reductio ad absurdum of (2), since each would then be an 'is' from which 'oughts' can obviously be inferred (If you want to eat ice cream, you ought to go to the ice cream shop; you want to eat ice cream; therefore you ought to go to the ice cream shop.) You've proven that atheistic naturalists (and, indeed, everyone else) should not accept (2), which is not obviously required by atheistic naturalism itself. If you take the other option and classify each of them as an 'ought', though, then there seems no reason why atheistic naturalism would rule all oughts out as part of the natural world. But if atheistic naturalism doesn't rule them out, then all the argument says is that you can't infer what you ought to refrain from doing from that part of the natural world that is classified under 'is'; it still leaves entirely open inferring it from that part of the natural world that is classified under 'ought'. The atheistic naturalist would escape through the loophole.
Puzzle 2: Suppose, however, you were to rework the argument so as to clarify Puzzle 1. There is still another puzzle. (5) and (6) don't give you (7). What they do give you is this:
(7') For any action, you are not obligated not to perform that action (i.e., it is permissible to perform it) if the obligation not to perform that action can only be reached by inferring it from something else.
That is, there is an additional assumption required to get to (7), that the only way you could get any ought (or at least an obligation not to do something) is by inferring it from something else. But mixing this hidden assumption with (2) means that there are no oughts unless there is an infinite regress of them. But if to get any ought we could only get it by inferring it from some other ought, that seems as much as to say we can't get any 'oughts' at all: it is to say that you have no oughts unless you infer them all from an infinite series of oughts, inferred from other oughts, that you already know. And this seems an odd position for anyone to accept, because it seems clearly contrary to the way we reason about oughts. The same two escapes seem to arise from this puzzle as arose from the other: the atheistic naturalist could then take the argument as a proof that he should not accept (2), or the atheistic naturalist could escape through the loophole by taking some ought as primitive, either because it is intuitively recognized or because it is constructed. And while one might raise problems with this loophole, they are problems far removed from anything touched on by this argument.
I think it should be recognized that there is a certain cleverness to the argument; one finds clearly stupid arguments on all sides of this dispute, and I don't think this particular one can be regarded as one of them. Its points of failure could easily be missed by just about anyone who wasn't thinking through the argument point by point. But it does fail to do what it is supposed to do, and, what is more, the failures seem to be fatal: there seems to be no way to revise this particular argument in order to avoid them. Even at its best it could not do what it is supposed to do; all it could show is that either the atheistic naturalist should reject (2) in any unqualified form, or that he should admit that there are oughts in the natural world, or that he should reject that all oughts must be inferred. But there are already good reasons for taking the first option; there are already possible candidates for reasons why one might take the second option; and there is at least one good reason for taking the third option if (2) is still accepted, to wit, that combined with (2) it already requires us to say there are no oughts at all, withough even bringing atheistic naturalism into the picture.
And that third point is what especially seems to deal the devastating blow to the viability of this particular line of argument: if the argument were accepted, the principles of the argument would seem to make (7) true for any possible position. It's true that the argument raises a problem severe enough that, if its steps are taken all to be in order, we should reject one of the premises; but rejecting the first premise won't solve the problem. The bleak conclusions can't be blamed on (1) in particular, even if (1) is false.