Yes I can—No you Kant—

Spitballer Dr. Johnson, sticking out his tongue, refutes the idealism of Bishop Berkeley. Drawing by Godescalc. Story at Frank's place (note that philosophically speaking Boswell was right and Johnson was wrong).

A seriously annoying op-ed in today's Times by Vlad Chituc and Paul Henne boasting that they have found empirical evidence that disproves an axiom from Emmanuel Kant, which is like saying you can disprove Euclid by baking a lasagna.

The axiom is the one leading up to the categorical imperative that "ought" implies "can", or:

one cannot be obligated to do something unless the thing in question is doable. For instance, there is no sense in which I am obligated to single-handedly solve global poverty, because it is not within my power to do so.

Apparently some analytical philosopher called Walter Sinnott-Armstrong contested that axiom in a celebrated paper of 1984 by the usual analytic method of combining a wild Talmudic imagination with a certain English linguistic prissiness, by constructing cases in which it would be possible to say that you are morally obliged to do something you can't do; as Chituc and Henne exemplify it,

Suppose that you and a friend are both up for the same job in another city. She interviewed last weekend, and your flight for the interview is this evening. Your car is in the shop, though, so your friend promises to drive you to the airport. But on the way, her car breaks down — the gas tank is leaking — so you miss your flight and don’t get the job.

Would it make any sense to tell your friend, stranded at the side of the road, that she ought to drive you to the airport? The answer seems to be an obvious no (after all, she can’t drive you), and most philosophers treat this as all the confirmation they need for the principle.

Suppose, however, that the situation is slightly different. What if your friend intentionally punctures her own gas tank to make sure that you miss the flight and she gets the job? In this case, it makes perfect sense to insist that your friend still has an obligation to drive you to the airport. In other words, we might indeed say that someone ought to do what she can’t — if we’re blaming her.

So what Chituc and Henne did was to submit this kind of question to a sample of "hundreds" of respondents (presumably including no more than the normal distribution of analytic philosophers), and found they could get a majority (60%) to agree that the sabotaging friend "ought" to keep the promise even though it was physically unkeepable, while under a third (31%) thought the original blameless friend "ought" to do it; therefore, they argue, ability to perform your obligation is a less important factor than blameworthiness. We "ought" to do something if we'll get blamed for not doing it, regardless of whether we are able to do it or not.

This is so stupid I can't stand it. In the first place, the idea of doing ethical philosophy by majority vote betrays a total misunderstanding of what philosophy is. It's not a game of "Family Feud". The most the questionnaire study can prove is that if Sinnott-Armstrong is right, then pretty substantial majorities of the population are right along with him, but (1) the majority could be wrong, you know, this has happened once in a while in the course of human history, and (2) the questionnaire could be poorly constructed, distorting the results.

And in the second place the questionnaire is poorly constructed. I don't know why it is that what is sometimes called "linguistic philosophy" is practiced so much by people who don't know any linguistics, but there's an elementary error in the grammar of the setup story: it uses the wrong verb. Namely, in the above example, it is not the case that

(a) she ought to drive you to the airport

with a punctual infinitive, but that

(b) she ought to have driven you to the airport

with a perfective. Where she should be now is not so much doing the right thing as basking in the sense of having done it. She is a moral failure now not because of what she isn't able to do now, but because of what she failed to do, or arranged to be unable to do, earlier, when it was possible. When she acquired the obligation, she could have carried it out, and the fact that she no longer can doesn't change her moral status. But the incorrect formulation in (a) parleys a grammatical confusion into a moral one and smushes the point into unrecognizability.

This shows why the Sinnott-Armstrong objection is a silly red herring. The obligation comes into existence together with the possibility, but may be said to persist after the possibility has evaporated, when it's too late to fulfill it but not too late to feel that one has failed to do the right thing. The blameworthiness is a secondary and squishy factor (on a scale from real innocence in the face of pure bad luck, where she still feels obliged to apologize but you have to assure her that it's not her fault, to real guilt, where you're never speaking to her again).

If they had asked their respondents with the formulation in (b) they would have gotten far more interpretable results, with more like 90% agreeing with the statement applied to the saboteuse, and maybe 30% agreeing with it as applied to the victim of bad luck (because of the ambiguity of "ought", which usually has a deontic sense, referring to moral obligation, as in "you ought to be kind to the poor," but sometimes an epistemic sense, referring to probability, as in "it ought to be raining this afternoon." But that would take it right out of the realm of testing Kant's idea.

Irving Berlin, taking cute to a realm of real greatness.

Sorry for the self-indulgence. There is a political application, though, as in:

(c) Obama ought to put a public option in his health insurance program.

(d) Obama ought to have put a public option in his health insurance program.

Sentence (c) is self-evidently stupid ("And ask him to bring me a beer as long as he's in the kitchen, too"), and (d) is true, but the question of whether he's to blame or not is arguable and logically independent.