Here's a thought inspired by anankastic conditionals. "Must" does not always imply "ought" in anankastic conditionals that start "If you want to...."
Background (lifted from a French farce I will not name): Pierre is late for a plane. The plane will leave without him unless he phones in a bomb threat.
Consider these said to Pierre:
(1) If you want to catch the plane, you must[/you'll have to] phone in a bomb threat. (2) If you want to catch the plane, you ought to phone in a bomb threat.
It strikes me that (1) is true and (2) is false.
The proposal that Kai von Fintel and Sabine Iatridou give for anankastic conditionals accounts for the truth of (1) fine. Essentially, they promote the goal of catching the plane ahead of any other goals the subject has, and evaluate possible worlds according to that goal (where the possible worlds are constrained by the modal background); if in every world in which that goal is fulfilled, p is true, then "if you want to catch the plane, must(p)" is true. That's true here--given the background, in every world in which Pierre catches the plane, he phones in the bomb threat.
But their proposal for "If you want to catch the plane, ought(p)" is essentially to promote catching the plane before all other goals, look at the possible worlds in which that goal is fulfilled, and see if p is true in all the best of those worlds, where "best" is evaluated with respect to the original ordering source plus the new goal. That predicts that (2) should be true--in fact it predicts that "If you want to A, must(p)" entails "If you want to A, ought(p)." If p is true in all the A-worlds, it's true in all the best A-worlds.
People like me who believe in moral dilemmas won't have a problem with the failure of must to imply ought--if you have conflicting obligations, it may be that you must do something that you oughtn't to do. But here there's no moral dilemma. Pierre's putative goal of catching the plane--and this may not be a goal that he even has--is in conflict with his duty not to call in bomb threats. So even if he wants to catch the plane, he ought not to take the necessary means to that end.
It seems to me that in (2) the goal of catching the plane is evaluated as more important than just an ordinary goal, but not so important that it overrides all other considerations. To say more we might have to come up with a theory of practical reasoning that says when commitments you have taken on should be overridden. I will come up with that theory in my next post [um, irony].
Posted by Matt Weiner at March 16, 2004 07:08 PMI think I agree with your observation about (1) and (2). So, is it that "ought" necessarily comes with a moral dimension (perhaps on top of whatever else is going on, such as giving advice on how to achieve a goal)? What are your intuitions about "should":
(3) If you want to catch the plane, you should phone in a bomb threat.
Anyway, I'll be thinking about the puzzle with (2). Thanks.
Posted by: Kai von Fintel at March 16, 2004 08:39 PMMy intuition is that it needn't be moral--the same situation can arise when the necessary means to the designated goal is not immoral but stupid.
If the immorality/stupidity is inherent in the designated goal, rather than just necessary under the circumstances, it seems to me that this doesn't arise:
(4)If you want to delay the plane, you ought to phone in a bomb threat
seems true.
I'm not sure about (3); my first instinct was to say that "should" is just like "ought," but (3) seems better than (2) at the moment.
Posted by: Matt Weiner at March 16, 2004 11:18 PMOh, and these seem acceptable to me:
(5)If you really want to catch the plane, you ought to phone in a bomb threat
(6)If you must catch the plane, you ought to phone in a bomb threat.
Well, now that I type it (6) seems somewhat dubious. I think it is predicted to be true by the semantics I suggested for "if you must." And in (5) the "really" seems as though it might mean that the designated goal is promoted even more than in (2), perhaps enough to override the reasons for not phoning in the threat; so that would be compatible with the analysis I wave at in the last paragraph of this post, I think.
Posted by: Matt Weiner at March 16, 2004 11:23 PM