Philosophy, Film, Politics, Etc.

Sunday, May 15, 2011

Devitt Comes To Szczecin

This week I had the pleasure of meeting and talking with Distinguished Professor Michael Devitt (CUNY). On Thursday he gave a talk at the University of Szczecin, and on Saturday there was a full-day workshop focusing on his current work in the philosophy of language. In between, on Friday night, he socialized with local philosophy students over beer and wine (and food). I was fortunately able to participate in all of these events.

Devitt and I are both naturalists and atheists. We are both sympathetic to ordinary language philosophy and Ryle, and we both have big problems with Stanley & Williamson's 2001 paper, "Knowing How." Unfortunately, however, I don't think Devitt is Rylean enough.

I see talk of minds and mental states just as talk of complex and indefinitely heterogeneous dispositions. Devitt sees it as talk about functional structures which supervene on neurological states. Devitt is strongly anti-behaviorist, favoring the Representational Theory of Mind (and with a sympathetic ear to the Language of Thought hypothesis), though he says he's amazed at how ingenious defenses of behaviorism can be. I'm sympathetic to behaviorism and amazed at how weak the arguments against it tend to be.

This difference might explain a disagreement we had during the workshop on Saturday, when I asked what support he had for his claim that any theory of linguistic communication must start with an analysis of the representational properties of linguistic terms. To me, that's just wrong. It's like saying any theory of team sports must start with an analysis of some particular properties of teams. Linguistic terms may have representational properties, but such properties are not required for them to be used in communication. Even when an entity has representational properties, those are not necessarily their only or most important features. At least, that's my view. So I asked Michael why he makes that first step, claiming representational properties are the most important and fundamental elements of a theory of linguistic communication. His response was a verbose variation on the incredulous stare. How could it be otherwise? he asked. He cannot imagine how else such a theory could begin. So I explained my behaviorist alternative. I pointed out that even in his own examples, what is first observed and analyzed is behavior, and that representational properties are only postulated if they help us explain the behavior. He seemed to simultaneously accept this point and reject it. Somehow. Maybe it's more accurate to say he dismissed it and moved on.

I raised some other points during the workshop, such as the fact that he does not properly motivate a theoretical distinction between what is said and what is meant. He argues heavily against the tendency to trust folk intuitions, saying that we must find independent, theoretical motivations for our distinctions. So while we both agree that there is a useful folk distinction between what is said and what is meant, I don't think it is theoretically interesting. Devitt disagrees, but his argument is too weak.

On Thursday and Friday, we discussed some issues concerning his recent paper, "Methodology and the Nature of Knowing How," which he presented on Thursday.

The first and biggest problem is that Devitt misinterprets Stanley & Williamson (S&W). This is partly explainable by the fact that S&W misrepresent Ryle. Devitt rightly interprets Ryle's knowing-how/knowing-that distinction more or less along the lines adopted by cognitive scientists, psychologists, and the like: Knowing-that is declarative knowledge while knowing-how is procedural knowledge. Devitt isn't quite so clear on what this distinction entails, but he suggests that knowing-that alone entails conscious and explicit representations. It is therefore dependent upon language use. Procedural knowledge is not. In fact, in Devitt's view (which I agree with), language use depends on non-propositional linguistic competence, which we may call a variety of knowledge-how. So knowing-that depends upon and cannot be reduced to knowing-how, just as Ryle says.

The problem is, S&W do not identify knowing-that as declarative knowledge. They call it "a relation between a thinker and a true proposition." They do not suppose it requires explicit, conscious representations or even linguistic articulability. Jason Stanley says (in correspondence) that it is just silly, even idiotic, to suppose that propositional knowledge must be articulable in a language. It's thus clear that Stanley's (and S&W's) distinction between knowing-how and knowing-that is not the distinction between declarative knowledge and procedural knowledge. But this undermines Devitt's entire line of argument. His criticism is that they challenge the empirically proven declarative/procedural distinction on purely linguistic grounds. They do no such thing.

I pointed this out to Devitt, and he seemed taken aback. He seemed to have no idea what sort of view of propositional knowledge they could have in mind. How could propositional knowledge be inexpressible, he asked? I pointed out that S&W make no metaphysical commitments, and that they say knowing-how can exist solely in virtue of having complex sets of dispositions. This didn't sit well. Devitt sees no reason to call procedural knowledge "propositional." Devitt would rather abandon talk of propositions altogether, in fact. We can talk about declarative knowledge and procedural knowledge, so far as those are useful for scientists. So he is willing to call declarative knowledge "propositional." Any other notion of "propositional knowledge" is more trouble than its worth. I have made a similar argument, in fact. In any case, while this might amount to a legitimate criticism of S&W, it is not the criticism Devitt makes in his paper.

Another issue I raised was with Devitt's discussion of a paper by Bengson, Moffett and Wright (BM&W) entitled "Folk Intellectualism," which tests Alva Noe's prediction that the folk will not attribute knowledge how to X to a person who has never been able to X. Noe is wrong, as it turns out. For example, imagine an accomplished skier and successful ski instructor who teaches people how to perform complex stunts, but who has never successfully executed any of them herself. The folk show a very strong tendency to attribute knowledge how to do these stunts, even though the instructor has never been able to do them.

Devitt thinks this research counts as a mark against Ryle. That is mistaken. BM&W's results contradict Noe, but they have no implications for Ryle. BM&W even acknowledge that Ryle's view of knowing-how is more complex than the one they are considering, which they prefer to call "Neo-Ryleanism."

BM&W make a mistake, though. They correctly suppose that knowing-how involves understanding X. Ryle makes the same observation: Understanding is part of knowing how, and this can be by virtue of any number of complex, indefinitely heterogeneous dispositions, and not simply the ability to X. However, BM&W wrongly conclude that they have motivated radical intellectualism, the view that knowing-how just is a case of knowing-that. They suggest that knowing-how is a combination of knowing-that-X and understanding X. Perhaps sometimes knowing-how involves both propositional knowledge and understanding. However, knowing how need not involve propositional knowledge at all, a point which BM&W have not refuted. So their intellectualist conclusion lacks motivation.

Getting back to Devitt, he makes one more mistake with this BM&W business. During his talk, he misrepresented BM&W's research, saying that the ski instructor was able to instruct people merely by giving verbal descriptions, and that the ski instructor could not ski at all. Yet, as I noted above, the instructor in their scenario is an accomplished skier and the nature of the instruction is not specified. In his paper he says that BM&W show that the folk attribute know-how to a person who can merely give a full description of a performance without being able to carry out the performance. This is not true. They do not say the instruction is purely via verbal descriptions. They do not draw the conclusion he says. They only conclude that a person who cannot do a complex stunt successfully, but who instructs a person to do that stunt successfully, can be said to know how to X.

When I pointed this out, Devitt asked, "What's the difference?" What's the difference between a person instructing by giving a full description and instructing in some other way? I think the answer is obvious. A computer can provide any finite description of a behavior, but we would not say the computer can thereby instruct. Instruction requires interaction, where the instructor observes and responds intelligently. The right sort of computer could be a ski instructor, of course, but not simply by giving descriptions. It has to interact intelligently. That's the Rylean point. Devitt didn't accept it, though, and frankly seemed a little unhappy about it. He rather abruptly changed the subject.