Ought and Can: What Implies the Implication?

The foremost question posed by libertarians to compatibilists is how a thoroughgoing determinism can leave room for moral responsibility. Another significant question, however, is how it can leave room for moral obligation, and what sort of obligation would be implied. One way to approach this question is through Frankfurt's rejection of the principle of alternative possiblities (PAP) as an obstacle in the way of making responsibility compatible with determinism. As David Widerker notes (1), it remains an open question whether Frankfurt’s rejection of (PAP) requires “renouncing the Kantian thesis that ‘ought’ implies ‘can’, i.e.

(K) An agent S has a moral obligation to perform (not to perform) an act A only if S has it within his power to perform (not to perform) A.”

Whatever the virtues of this debate, it is actually very clear that Frankfurt cannot accept the Kantian thesis (though he claims he can) for the simple reason that (K) is very obviously not a correct representation of that thesis. No understanding of the English (or German) language, however charitable the interpretation, is sufficient to yield (K) as an accurate restatement of the claim that “ought” implies “can.” Instead, as the wording of the implication clearly implies, the thesis is:

(K1) If an agent S has a moral obligation to perform (not to perform) an act A, then S has it within his power to perform (not to perform) A.

Of course the two are logically equivalent, so that (K1) entails (K): if we can establish on independent grounds that S cannot perform (not perform) A, then we cannot reasonably say that S has an obligation to perform (not perform) A. For example, if there were a species of beings that were rational but literally could not avoid constantly and intentionally killing each other (so, only a little different from human beings), then they could not have an obligation not to kill each other. So doesn’t this just mean that (K1) really boils down to (K)? I don’t think so. A Kantian might question the coherence of any such thought experiment, and do so on the following Kantian grounds: to be under an obligation is to be a rational being for which reason is practical. But if reason is practical, then it follows that we are able to act on it. (K) and (K1) differ precisely in their emphasis: Kant is not trying to derive obligation from ability (an “ought” from an “is”), but to show that obligation has implications for how we must see ourselves; more specifically, for what abilities we must attribute to ourselves regardless of lack of any possible other evidence for them. This, obviously, is a very different project. Logical equivalency does not guarantee identity of meaning.

(K1) is intimately tied to a project of giving what may be seen as a libertarian spin to freedom: it is determined entirely on the basis of our obligations, and established despite the truth of determinism (which also makes it completely different from any libertarian theories around today that I am aware of). Thus we come to Frankfurt, who happens to be a compatibilist. As a compatibilist, he can accept (K). There is nothing incoherent about accepting both that (1) our abilities place limits on our obligations and (2) the extent of our abilities can be fully ascertained within the scope of the natural sciences (whether deterministic or not). But (K1) meshes only with (1); it implies—strictly implies—that (2) is at least possibly false. It follows that Frankfurt would have to reject (K1). Would he have to do it on the basis of his rejection of (PAP)? I don't see how, except insofar as his rejection of (PAP) comes with other compatibilist assumptions.

(1) “Frankfurt on ‘Ought implies Can’ and Alternative Possibilities” in Analysis, Vol. 51, No. 4. (Oct., 1991), p. 223.