Welcome to our first Oxford Studies discussion, on John Brunero‘s “Cognitivism about Practical Rationality” (in OSME 9). The essay is available through this link for a few months, courtesy of OUP. Kieran Setiya has kindly contributed a commentary, below the fold. Please join us!
Kieran Setiya, August 19, 2015
Over the last ten years, John Brunero has written a series of trenchant, illuminating papers on the rationality of intention, many of which deal with problems of instrumental reason. His latest paper is no exception: it is a powerful development and critique of the “cognitivist” approach to instrumental rationality, on which the requirement of Means-End Coherence follows from requirements of theoretical reason on our beliefs. As someone who is sympathetic to this approach, I welcome the development but resist the critique.
John’s topic is the following principle:
Means-End Coherence: Rationality requires that [if one intends to E and believes that one will E only if one intends to M, then one intends to M].
He considers, first, an attempt to explain this principle by deriving it from
Closure: Rationality requires that [if one believes that P, and that if P then Q, then one believes that Q]
together with the Strong Belief Thesis, that intending to X involves believing that one will X, and the further claim that it is irrational to have false beliefs about one’s own intentions.
The initial problem turns on counterexamples to Strong Belief, as when I intend to lift a heavy log but am doubtful that I’ll succeed. Some respond by denying that I have this intention: at most I intend to try. John concedes this response, for the sake of argument, but doubts that it is of use to the cognitivist. He presents the following case:
I want to lift a heavy log. I am doubtful that I will succeed, but I intend to try. I believe that I will lift the log only if I intend to bend my knees. But I do not believe that intending to bend my knees is necessary for trying to lift the log, since I tried (and failed) to lift it last time without any such intention. I do not intend to bend my knees.
According to John, I am means-end incoherent, even though I do not violate Means-End Coherence on the Strong Belief account. The cognitivist cannot explain where the incoherence lies.
I am sympathetic to this objection, though I hesitate a bit. In general, I can try to lift a log without intending to bend my knees. No problem there. But it is possible to do so when I know I won’t succeed without that intention? More importantly, is it rational to believe that I will lift the log only if I intend to bend my knees without believing that I will try to lift the log only if I so intend?
Can I regard what I am doing as a genuine attempt to lift the log when I believe it is bound to fail? If not, the cognitivist can extend her treatment to the case above. (It does not matter that I tried to lift the log last time – before I knew that bending my knees was necessary to lift it – without intending to bend my knees. My situation is now different.)
These remarks are speculative, and I do not place much weight on them. In the end, I agree with John that cognitivists should operate without Strong Belief. As I have argued elsewhere, we should think of the will as a capacity for practical knowledge, a capacity that issues in knowledge of what I am doing, or what I am going to do, when conditions are conducive. In the presence of interfering factors, lack of know-how, and so on, the will may issue instead in practical confidence or partial belief, which plays a similar role in motivating action.
I am thus encouraged by John’s attempt to explain Means-End Coherence on weaker terms. If intending to X involves some degree of belief that one will X, this vindicates the
Very Weak Belief Thesis: Intending to X involves not believing that one certainly will not X.
John argues that, if an agent’s instrumental belief takes the following form:
If I do not intend to M, I certainly will not E.
and she both satisfies Closure and intends to E, she cannot believe that she does not intend to M. Assuming that it is irrational not to believe that one does not intend to M when one in fact does not, it follows that, if she is rational, the agent intends to M, and so conforms to Means-End Coherence.
The steps of this argument are contentious. As John notes, the assumption of rational self-knowledge involved in it is very strong: not just that it is irrational to believe that I intend to M when I do not, but that it is irrational to be ignorant of the absence of intention. Another question concerns the interpretation of epistemic modals such as “certainly” in conditionals like the one above. But I am optimistic that the details can be worked out. If we shift from epistemic modals to the degrees of beliefs with which they are closely connected, the basic idea is this. If I intend to E, and believe that intending to M is a necessary means to E-ing, I have some degree of belief that I will E and I am certain that I will not E unless I intend to M. It is irrational to have these attitudes without having some degree of belief that I intend to M. If it is irrational to have such confidence unless I intend to M, it is irrational to violate Means-End Coherence.
John objects to this line of thought not on points of details, but in principle. It is here that I am less convinced. John makes two main claims, each of which I will briefly address.
First: cognitivism is unpromising as a general account of practical rationality. Once we give up Strong Belief, it is not clear how it explains the demand for consistency among intentions, let alone such requirements as “enkrasia” or instrumental rationality in relation to non-necessary means.
I mostly accept this. Unlike some cognitivists, I do not take cognitivism to be a general account of practical rationality. Instead, it explains why certain alleged requirements of practical reason are very different from the rest. I would say that Means-End Coherence is a requirement of theoretical rather than practical rationality. I think cognitivists can explain why it is irrational to intend to do A and intend not to, since that would involve conflicting shifts in practical confidence. But I am wary of more general demands for consistency in intention and belief. What is more, while reason must have something to say about preferences, values, and non-necessary means, it is not easy to formulate strict principles, like Means-End Coherence, that apply to them. Is it irrational not to intend the most effective means to one’s ends? Arguably not, since there may be reasons against them: efficiency is not the only concern. In general, I think of practical reason as the application of ethical virtue to practical thought. Means-End Coherence is something else, and cognitivism explains why.
Second: the more serious worry, for John, is that, on the cognitivist account, the agent who suffers from means-end incoherence – intending to E, believing that she will E only if she intends to M, but not intending to M – has only one way out. Her only rational response is to conclude that she will not E, and so to give up the corresponding intention. In contrast, he suggests, the requirement of Means-End Coherence is symmetric. This feature is “unexplained by the cognitivist account, and so the cognitivist account is inadequate.”
I think John is right about the implications of cognitivism, but they are implications we should embrace. As John notes, the proper formulation of Means-End Coherence appeals not to the belief that M-ing is a necessary means to E-ing, but to the belief that intending to M is necessary. More specifically, one must believe that one will E only if one now intends to M. Otherwise, it may be rational to intend to E without intending to M, as when I trust myself to form the intention later on. But then it is simply true that, if I violate Means-End Coherence, the only rational response is to conclude that I will not E. It is too late to form the intention to M: I believe that I will E only if I now intend to M, which I do not. Given Very Weak Belief, I must give up the intention to E.
The upshot is that there is no symmetry in how one should respond to means-end incoherence. One should give up one’s intention for the end. If there is symmetry, it lies in how one should adjust one’s attitudes to conform to Means-End Coherence in anticipation. When I intend to E and foresee that intending M will be necessary for doing E, I can prevent the threat of future incoherence either by giving up the intention to E or by forming the intention to M. But cognitivists allow for this: it is part of their view that one can rationally form intentions, and the degrees of belief that go with them, in ways that go against one’s prior evidence.
If this is right, John’s objections are inconclusive: his paper is not a refutation of the cognitivist project, but a valuable contribution to it. I would be happy to discover this, even if he is not!