Intelligent activity requires the use of various intellectual skills. While these skills are connected to knowledge, they should not be identified with knowledge. There are realistic examples where the skills in question come apart from knowledge. That is, there are realistic cases of knowledge without skill, and of skill without knowledge. Whether a person is intelligent depends, in part, on whether they have these skills. Whether a particular action is intelligent depends, in part, on whether it was produced by an exercise of skill. These claims promote a picture of intelligence that is in tension with a strongly intellectualist picture, though they are not in tension with a number of prominent claims recently made by intellectualists.
In recent work about know how, Rylean regress arguments have largely dropped out of focus. They play little role in the anti-intellectualist arguments of various kinds in the papers collected in Bengson and Moffett (2011). They are used as something like target practice by intellectualists like Jason Stanley (2011), who uses the first chapter of his book to dispose of them before getting onto the real business. And even Yuri Cath, who in other work has launched sharp critiques of intellectualism, has argued that the regress arguments for anti-intellectualism don’t work (Cath 2011, 2013). The majority view seems to be that Carl Ginet (1975) basically showed these arguments didn’t work, and it’s time to move onto other considerations for or against intellectualism.
I think this isn’t exactly right. In particular, I think regress arguments can be used to show a few different things. For one, they can be used to refute a precisification of this thesis, which plays a key role in some intellectualist arguments.
Once we have seen that thesis is false, we need a new picture of how action can be at once intelligent and non-volitional. Some considerations similar to those adduced by Ryle (1949) concerning agents who either concentrate on irrelevant considerations, or ignore relevant ones, show there is a role for intellectual skill that cannot be identified with any piece of knowledge that. And some further considerations, similar to those adduced by Cath (2011), suggest that this intellectual skill can’t even be constituted by a piece of knowledge that. So regress arguments, I’ll argue, can do quite a lot to motivate the thought that there was a lot wrong with the intellectual picture Ryle tried to attack.
The position I’m going to be defending is a long way from the strongest kinds of Rylean position that contemporary intellectualists such as Stanley are focussed on arguing against. My focus is primarily on intellectual skill. This has some relevance for debates about know how, though less relevance for debates about the semantics of know how ascriptions. This focus on skill rather than know how ascriptions is hardly novel; it is continuing a trend that we see exemplified in recent work by, inter alia, Carlotta Pavese (2013), Ellen Fridland (2014) and Cheng-Hung Tsai (2014). And in fact that conclusions I’ll draw here are, I think, similar to the ones that Fridland draws.
Once we move towards thinking about skill, we can get varieties of anti-intellectualism that are very different from those that were the focus of most philosophical discussion until very recently. For example, the anti-intellectualist view I’m defending is consistent with the following four theses.
Instances of intellectual skill are usually, and perhaps always, not happily reported using know how ascriptions.
Know how ascriptions are rarely, if ever, reports of intellectual skill, and are frequently reports of propositional knowledge.
Intellectual skill is guided by, and dependent on, propositional knowledge.
Propositional knowledge is not behaviourally inert.
Not just is the view consistent with these four, I’m fairly confident that the last three at least are true. But that’s all consistent with the view that intellectual skill is not itself propositional knowledge. And it’s all consistent with the view that we can learn philosophically significant conclusions from Ryle’s regress arguments.
One disclaimer before I start. Although this paper is heavily influenced by Ryle (1945, 1949), and sympathetic interpreters of Ryle such as Jennifer Hornsby (2011), I make no attempt at Ryle exegesis here. I think there’s a decent case to be made that Ryle was sympathetic to the position defended here, but I’m going to leave that debate for another day.
Define a volitional action to be one that is preceeded by a volition to perform that very action. And say an action is normatively assessable if it can properly be assessed using terms like praiseworthy, blameworthy, intelligent or stupid. Note that I’m ruling out assessments as good or bad as versions of normative assessment, in the relevant sense. Someone who has a good digestive system is not, thereby, normatively assessable in the stipulative sense I’m using. Both of these definitions are to an extent stipulative; the terms ‘volitional’ and ‘normative’ can sensibly receive many other definitions. Still, I will stick to these definitions here. In light of those stipulations, consider the following set of propositions.
Only volitional actions are normatively assessable.
The action of forming a volition is normatively assessable.
Some public actions, such as making a move in a chess game, are normatively assessable.
It should be obvious that this leads to a regress. Whether the kind of regress in question is impossible, or even impractical, is a tricky question. (See Robert K. Meyer (1987) for some of the complications that arise when trying to reason about regresses.) But it is commonly assumed in this literature that the kind of regress that these three premises lead to is problematic.
Since the third premise is obviously true, the issue is whether the first or second is false. But it seems that second is true as well. Just as we can assess a person’s actions as praise or blameworthy, intelligent or stupid, we can assess the process by which she decided to perform those actions in the same way. Consider two people who make the same, as it turns out great, chess move in the same situation. The first notices an initially appealing counter to her move, and sees after careful thought that it won’t work. The second simply doesn’t notice the counter, and is stumped when her opponent makes it. It seems the first has engaged in a more intelligent practice of volition formation than the second. Or imagine a third player, whose initial analysis of the move starts by considering a recipe for arroz con leche. Unless there turns out to be an unnoticed connection here, this looks even less intelligent than the second player.
On the other hand, the first premise is rather unintuitive. To borrow an example from Angela Smith (2005), it is blameworthy to forget a friend’s birthday, although forgettings are rarely volitional. So we must reject 1 or 2, and while 1 is subject to independent counterexample, 2 seems independently plausible. So 1 must be false.1
That’s already a substantial conclusion. Something like 1 is behind William Alston’s famous, and influential, arguments against deontological approaches to epistemology (Alston 1988). But the negation of 1 is not a novel claim; I’m saying nothing here that Smith didn’t say in her rejection of the “volitional view of responsibility” (Smith 2005, 238). And similar views have been put forward by other critics of Alston such as Sharon Ryan (2003) and Matthias Steup (2008).
But still, the fact that 1 is false seems not to have been sufficiently appreciated in the recent literature on intellectualism. To see one place where it is relevant, consider this set of propositions, which also seem to trigger a regress.
Intelligent action requires the triggering of a prior representation of knowledge relevant to the action.
The triggering of a representation, when done well, is an intelligent action.
Some public actions, such as making a move in a chess game, are intelligent.
Again, these propositions obviously trigger a regress, and that seems like good evidence to take one of them to be false. This is very similar to one of the regresses Jason Stanley considers in chapter 2 of his (2011). And Stanley thinks the false proposition is 5. He writes “Triggering a representation can be done poorly or well. But this does not show it can be done intelligently or stupidly.” (Stanley 2011, 16) Indeed, he writes that since “triggering representations is something we do automatically” (Stanley 2011, 16) a statement like 5 is a “manifest implausibility” (Stanley 2011, 16). But the argument here relies on 1. If you think things done non-volitionally can be intelligent or stupid, it isn’t too much of a stretch to think that things done automatically can be intelligent or stupid. Indeed, Smith’s birthday example is already enough to undermine Stanley’s point; forgetting a friend’s birthday seems automatic in the sense he has in mind, but is also stupid.
More generally, it seems very intuitive to describe everyday cases in such a way that 5 must be true. For example, Billy asks Suzy whether she thinks Jill’s party will be a success. There are a lot of things that are common knowledge between the two of them. One is that Jill is a proficient party host. Another is that Jill has invited all of their colleagues, including Jack. Another is that parties which Jack attends are rarely successes. But Suzy thinks for a minute, remembers that Jack is away in Ohio, and says that it will be a success.
It was smart of Suzy to think about Jack’s whereabouts. It wasn’t, perhaps, necessary. If she’d just reasoned from Jill’s general proficiency to the success of the party, she would have got to the right conclusion. But it was better to note a possible complication, and check that it wouldn’t actually get in the way.
It would have been stupid to perform the same activity for many other kinds of possible complications. If Suzy had thought to herself, “The party will be a disaster if there’s an alien invasion in the middle of it, but there’s no reason to think the aliens will invade just now, so I’ll keep on thinking it will be a success,” that would have been stupid. Other possible complications are not stupid to consider, but they are intellectual mistakes. The party won’t be a success if there’s a police raid in the middle, based on a mistaken view the police have about where a particular drug dealer lives. Police do make mistakes, so even if Jill isn’t a drug dealer, this could be a genuine concern, depending on how nearby the mistakes are. But if the nearest mistake was a botched raid in a neighbouring state in the previous year, it’s wrong for Suzy to worry about this before answering Billy’s question.
Stanley’s view has to be that I’ve been misusing adjectives systematically through the last two paragraphs. I shouldn’t have said that it was smart of Suzy to consider Jack’s whereabouts, or that it would have been stupid to consider the alien invasion. Rather, it was just her cognitive system working well when she considered Jack, and would have been working poorly had she considered the aliens, and sub-optimally had she considered the police. This doesn’t seem at all the natural way to describe the case to me, in part because I’m not sure I see the difference Stanley is hinting at. Intelligence just is the good operation of the cognitive system, and stupidity its poor operation.
So these two regresses lead to two interesting conclusions. First, some non-volitional actions are normatively assessable. Second, intelligent action does not always require the prior triggering of a representation of relevant knowledge. Both of these are interesting. Both of these are negations of part of what you might consider “the intellectualist picture.” (Cath (2013) notes that Ryle often refers to the regresses as arguments against this picture, not against any particular thesis.) But neither of them get us very close to a distinction between know how and know that, or between intellectual skill and know that. The next section addresses some ways we might move closer to arguments against more central intellectualist claims.
As noted in the introduction, my plan is not to offer an argument with regress like premises, and the conclusion that intellectual skill is distinct from propositional knowledge, or that know how is distinct from propositional knowledge. What I do want to do is sketch a picture of human intelligence (at a very high level of generality) that presupposes that intellectual skill is not identical to propositional knowledge, and suggest some considerations to the effect that no similarly plausible picture exists in which intellectual skill and propositional knowledge are identified. The thought here is not that the only way out of the regress involves distinguishing skill and knowledge – and perhaps distinguishing know how from know that – but rather that the best way out does.
Start with a well known, if not obviously authentic, exchange.2
Oscar Wilde: I wish I’d said that.
James McNeill Whistler: You will Oscar; you will.
Assuming this really happened, that’s a clever response. It’s an occupational hazard of philosophers to think that the ability to come up with quick, clever responses is somehow central to intelligence. But we can reject that wildly implausible view without thinking that it’s wrong to think of these quips as a manifestation of a kind of intelligence.
Now let’s think of how someone could have come up with this response. Even before we start researching the neural patterns behind quips like this, we can be pretty sure the following is not what happened in Whistler’s brain. He first made an exhaustive list of all possible responses, from “Green ideas sleep furiously” to what he actually said, then figured out which would be best, then produced the best one. On this wildly implausible model, the reply would be intelligent because it would reflect the speaker’s ability to properly evaluate this list of responses. That’s implausible because the list is simply too big. Indeed, it is in principle infinitely large. The list is too big to survey not just consciously, but subconsciously.
Coming up with a response like this requires first coming up with a narrower list of possible responses, and then evaluating which is best from that list.3 There’s a romantic model of intellect where the list in question consists of just the reply actually issued. On this model the perfect reply appears fully and perfectly formed in the mind of the intelligent person. Now such a model may often fit the phenomenology, but I don’t think we should give that much credit. It’s an empirical question how many possible replies are represented in the mind in a situation like this, before the chosen reply is issued. What’s not an empirical issue is whether the list of possible replies that is represented in the mind is finite or infinite. It simply must be finite, which means that there must be better and worse lists to consider. And that suggests that there is some skill involved in coming up with the list.
One could reject this last conclusion. One could try saying that the coming up with a list of possible replies is no manifestation of skill, but the skill is only involved in the evaluation and selection of replies. But this seems to generate a bizarre explanation about why the less skilled interlocutor comes up with worse replies. The model, presumably, is that the lack of skill does not explain having the wrong list of replies to choose between. Rather, what explains their less skilled reply must simply be that they misevaluated the possible replies. But that doesn’t fit the observed data. It’s much easier to see of someone else’s reply that it was clever than it is to come up with a clever reply.
It could also be objected that the model I’ve suggested is much too simple. It isn’t just that the mind issues a list of options, then evaluates them, and then selects the best. A more plausible model involves more recursive steps. The mind first generates a list of options, selects the best, then generates a list of refinements of that best option, selects the best of those, and so on. Perhaps when we consider superficial forms of intelligence, such as quips, it makes sense to consider a ‘one-step’ model, where a list is generated and evaluated, followed by a speech. But when one is choosing one’s words carefully, as in say Wilde’s writing, the simple model I’ve described feels much too simple.
But although the simple model is too simple for considered writing, the general structure must be right. Even a writer working at a leisurely pace, such as Joyce taking decades to write Finnegans Wake, doesn’t have time to consider, even subconsciously, all possible constructions. There are still too many. And nor is it true that the difference between Joyce’s skill and ours is that he realises the value of the sentences we all represent. The rest of us didn’t simply misjudge the value of “Nobirdy aviar soar anywing to eagle it” (Joyce 1939/2012, 505); we simply didn’t token it. The ability to token mental representations like that is part of what Joyce’s genius consists in.
I’ve focussed so far on cases where it is a priori implausible that human thinkers start by surveying the range of possible things they could do. It is also interesting to look at cases where this is in principle possible, but doesn’t seem to happen in practice. There have been, traditionally, major differences in the style of play between human and computer chess players. (Since so many young players learn from machines these days, Kasparov (2010) suggests these differences are diminishing.4) This isn’t necessarily because humans can’t consider all options on the chess board. Usually there will be fewer than a hundred available moves, and a human could consider each. But that isn’t, it seems, how humans think. They don’t allocate equal resources to working through each of the possible options. As a result, computers often come up with surprising kinds of moves. Now computers are actually very good at chess, so these pre-deliberative allocations of cognitive resources may not have been optimal. Perhaps it would have been better for traditional chess players to spend more time thinking through unlikely progressions of the game. But it is evidence that even when we could use an unintelligent method for beginning inquiry, namely recursively generating the possible options, we prefer to use intelligent methods.
So intelligent action, at least in humans in the kinds of situations humans normally find themselves in, consists in part of making intelligent choices about where to start inquiry. Given that intelligent action need not be volitional, as we established above, it isn’t surprising that being intelligent consists in part in starting in the right places. But perhaps this intelligence is just itself a kind of knowledge. It is, we might suspect, just the knowledge of what a good starting point will be. Or, since we will want to start with all and only the considerations relevant to a given inquiry, it is just knowledge of what is relevant.
The resulting picture is both perfectly intellectualist, and immune to the regresses considered above. The intelligent person knows what is relevant to what inquiry. Her choice of starting points is guided by this knowledge. (The ‘guidance’ metaphor recurs frequently in Stanley’s work.) This isn’t because it leads to a volition to start just here. Such a volition would be self-defeating, since in the relevant sense of ‘start,’ by the time this volition is formed, one has already started, and indeed started elsewhere. Nor can she be guided by even a triggered representation of this knowledge of relevance. Again, if that happens, she is in the relevant sense starting elsewhere. But perhaps propositional knowledge can guide directly; not by generating volitions, and without even being represented anew.
Now I don’t think this picture is right. But it isn’t incoherent either, and it takes work to see why it isn’t right.
One bad argument against this picture starts with the idea that skills are active, while knowledge is passive. The thought is that the person who knows a lot is like the Tortoise in Louis Carroll’s dialogue (Carroll 1895), only able to add more premises but never to reach a conclusion. It is only with skill that we can get to the conclusion. Stanley rightly objects to this argument on the grounds that it just isn’t true that knowledge is passive in the relevant sense. We should not, as Stanley puts it, “over-intellectualize knowing that.” (Stanley 2012, 773). (A similar point is made in Stalnaker (2012).) Knowing that p is not just a matter of having p written in a knowledge box somewhere in the brain; it can in part be constituted by active dispositions.
A better argument looks at the very different modal profiles of intellectual skill and knowledge of relevance. Someone can know that something is irrelevant and yet lack the skill to ignore it; or they can know that something is relevant and yet lack the skill to consider it in a timely manner. Examples from the other direction, where there is skill without knowledge, are a little more contentious, but we’ll look at some possible cases of those too. But first we’ll run through two examples to show how easy it is to have knowledge without skill.
Alice has spent a lot of money on video-conferencing equipment. But it isn’t working at all well, and she now has to decide whether to try and patch it into something better, or buy a whole new system. She knows the sunk cost fallacy is a fallacy; that buying a new system would make the previous purchases a waste is no reason to not buy a new system, especially if doing so is good value compared to the cost of buying a ‘patch.’ But she can’t bring herself to ignore this fact when deliberating. Even though she eventually makes the right decision and buys new equipment, she takes much longer about this than she would have if, say, the existing equipment was old enough that she could easily conceptualise it as obsolete.
Bob is trying to solve a puzzle about the properties of functions from rationals to rationals. He knows that it is often helpful, when solving such puzzles, to transform the puzzle into one about functions from ordered pairs of integers to ordered pairs of integers. He knows that in the sense that if you asked him whether it could be useful to consider that transformation of the puzzle, he would immediately say yes, and this answer would come with the phenemenology of recollection, not of new insight. But no one does ask him that question, and the transformation in question simply never occurs to Bob. Since the untransformed puzzle is very hard, while the transformed puzzle is manageable, Bob never solves the problem.
It seems to me that what’s happened in both cases is that the agent has some knowledge, but is incapable of using it. What they lack is a skill. In particular, they lack what Fridland (2014, 2746) calls ‘selective, top-down, automatic attention.’ Alice keeps attending to something she should not, even though she knows she should not. Bob fails to attend to something he should, although in some sense he knows that is what he should attend to. Bob’s case is one of the reasons I find the picture of skill presented by Stanley and Williamson (2017) unhelpful. They say skill is a disposition to form knowledge. But Bob has the important knowledge. The disposition he lacks is the disposition to activate that knowledge, and let it guide deliberation. That’s what constitutes his lack of skill.
It’s true that knowledge isn’t completely passive. If Alice never appealed to the fact that the sunk cost fallacy is a fallacy in her reasoning, we wouldn’t say that she knows it. If none of Bob’s answers were guided by the existence of natural and useful transformations between rational numbers and ordered pairs of integers, we wouldn’t say he knows such transformations are natural and useful. I’m here agreeing with Stanley and Stalnaker that knowledge is itself a kind of disposition. And intellectual skill is a kind of disposition too. But they are very different dispositions. In particular, they have very different triggering conditions. Bob lacks some skill because he does not call to mind this fact about rational numbers right now. He has the salient knowledge about rational numbers because he is disposed to use the facts in question often enough.
So intellectual skill and knowledge of relevance have different manifestation conditions, and so they are not identical. But we can say something stronger than that. The cases of Alice and Bob are not in any way unusual. Examples where we forget the salience of some consideration, or can’t get an irrelevant point out of our heads, are frequent. In principle, one could respond to the arguments I’ve made so far by saying that while knowledge of relevance is not identical to skill, nevertheless the two are as closely linked as, say, a material object and the matter that constitutes it. And if I had to resort to bizarre cases of the kind we torture introductory students with to make my point, I’d say that would be the right response. But given how normal Alice and Bob’s cases are, this seems like the wrong move. Skill and knowledge don’t just come apart in theory, they come apart in practice, frequently.
So far I’ve defended three theses that are in tension with some forms of intellectualism. They are:
Some non-volitional actions are normatively assessable.
Not all intelligent action is preceded by the triggering of representations of relevant knowledge.
Intellectual skill, in particular the intellectual skill associated with starting inquiry in the right place, is not identical to any piece of propositional knowledge.
While this doesn’t show that, for instance, know how and know that are distinct, and is completely silent on what we should say about know how ascriptions, it does undermine some intellectualist programs. I’ll conclude with some objections either to the arguments I’ve put forward, or to their significance.
Objection: Even if all of this is true, there may still be a sense in which intellectualism is true. After all, it could still be that knowledge guides action in a suitable way. (Compare (Stanley 2011, 2).)
Reply: This could be true. Whether it is a win for intellectualism depends a bit on the boring question of how we settle the term ‘intellectualism,’ and a bit on more interesting questions about priority. Let’s start by distinguishing five theories we might call intellectualist.
- The possession of an intellectual skill just is the possession of a piece of knowledge.
- The possession of an intellectual skill is, always, constituted by a piece of knowledge.
- The possession of an intellectual skill is, often, constituted by a piece of knowledge.
- The possession of an intellectual skill is, always, caused by the possession of a piece of knowledge.
- The possession of an intellectual skill is, often, caused by the possession of a piece of knowledge.
This paper has been arguing against Identity Intellectualism. I think the falsity of this is as much as we could reasonably hope to prove using regress arguments. (I think I’m here agreeing with Wiggins (2009) and Hornsby (2011).) And the considerations behind the regress argument do, I think, show it to be false. If someone wants to insist that by intellectualism, they mean something weaker than this, I’m not going to quarrel over terminology. I’ll just note that Identity Intellectualism is an interesting, and false, thesis.
The arguments here are clearly not arguments against either form of Weak Intellectualism. Indeed, they are naturally understood as the kind of cases that confirm Weak Intellectualism. Mathematics students, like Bob, train by learning a lot of mathematical facts. And it’s hard to see how they could develop the relevant skills without knowing some important facts. This is, I suspect, the general case. Skillfully bringing the right considerations to bear on a problem requires, and is probably the causal consequence of, knowing a lot of relevant facts. (Tsai (2014) makes clear how one can simultaneously hold that skills are in part constituted by knowledge of facts without having an intellectualist picture of skill.)
But what of the other two intellectualist theories? Do we have reason to think that there are some skills that are not constituted by, or not caused by, the possession of factual knowledge? One way to quickly show that would be to show that there can be skills without the related knowledge. Perhaps that’s not just sufficient for rejecting Constitutive/Causal Intellectualism, but necessary. If knowledge without skills is possible, as in Alice and Bob’s cases, and skills without knowledge were impossible, that asymmetry would call out for explanation. And something in the vicinity of Constitutive or Causal Intellectualism would be a very good candidate explanation.
There are (at least) two promising routes to showing that there can be skills without knowledge. One is due to Imogen Dickie (2012). She argues that since there are so many different routes to skill than there are to knowledge, we should expect that there will be cases of skill that are causally prior to knowledge. Jason Stanley (2012) replies that Dickie’s argument assumes an overly narrow conception of propositional knowledge. This is a fascinating debate, but I don’t have anything useful to add to it, so I’ll just note the existence of this route, and move on.
The other route is due to Yuri Cath (2011). He suggests that facts in virtue of which a person might lose propositional knowledge do not always bring about a loss of knowledge that. I’m going to sketch a Cath-style argument that we can have intellectual skills without knowledge. I think the argument has some force, though there are more ways to resist it than there are to resist the argument against Identity Intellectualism.
Ross and Rachel are economics students taking an exam. They are given a hard question asking about the likely effects of an exogenous shock, say an earthquake affecting an area the supplies crucial raw materials, on some related markets. The question is hard, with the relevant causal pathways being interconnected and often opposing. The only plausible way forward is to use a model and search for equilibrium points in the model. That’s what Ross and Rachel have both been taught to do. And in fact both of them quickly select the right kind of model, with just the right amount of complexity in it to answer the question without being overburdened, and set out on the difficult algebra involved in solving the question.
So far it looks like both Ross and Rachel have shown intellectual skill. Now it turns out Ross and Rachel have very different views about the role of models in economic thinking. (My own thinking about models has been heavily influenced by (Strevens 2008 ch. 8) and Davey (2011), and I rely on their insights in what follows.) These models involve, as all models do, some serious idealisation. Most notably, they assume that all the relevant actors are perfectly rational utility maximisers. Rachel hasn’t given much thought to this assumption, though she knows it to be literally false. But if pressed, she would say some reasonably sensible things about why she was using the model. For one thing, the familiar failures of human rationality aren’t obviously relevant to the puzzle being presented. For another, they’ve been taught that using these models is a good way to solve problems, and that testimonial evidence carries some weight. And for another, it’s an exam, and it is likely that questions have been selected to test how well students can use the models they have been taught. If those are her background, implicit, views, I think it is plausible to say that Rachel knows that the model is relevant to the exam question, even if she couldn’t produce a theory of idealisations in economics of the standards of the best philosophers.
Ross’s views about models are rather different. He thinks the familiar models in economics work, when they do, because the background assumptions are strictly and literally true. He thinks economic agents are utility maximisers, and the apparent evidence to the contrary is due to sloppy experimental design. He thinks markets are always in general equilibrium. And so he thinks that the only sources of error in predictions we can make about markets are from errors about things like the costs of extracting raw materials after the earthquake. This perspective is, of course, grossly mistaken. Moreover, Ross thinks that if the assumptions were not correct, there would be no point in using the models. This too is a mistake, though perhaps not as dramatic as his other mistakes.
Now even if Ross and Rachel aren’t thinking about these philosophical views about the nature of models, I think they are relevant to whether each of them know that the models are relevant to the puzzle. In particular, I think Rachel does know that the models are relevant, while Ross’s belief that they are relevant is more like a lucky guess than a piece of knowledge. Still, I think we should say that Ross showed skill in using this model rather than a more or less complex model, or a different kind of model, or no model at all. So he is a case of intellectual skill without knowledge of relevance.
I don’t think this case is conclusive. I can think of at least four ways someone might reasonably object to the case.
It might be argued that despite his false views about why the models are relevant, he really does know that they are relevant. In other words, we would have another counterexample, to be added to those discussed by Warfield (2005) and Luzzi (2010), to the theory that false beliefs cannot generate knowledge.
It might be argued that Ross is not really skilled, since it is a matter of luck that the falsity of his beliefs does not lead him to false conclusions here.
It might be argued that although Ross doesn’t know that this model is relevant, his skill is constituted by, or caused by, some other knowledge he has.
It might be argued that the broad picture of the role of idealisations in scientific reasoning that I’m adopting from Strevens and Davey is mistaken, and this fatally undermines my use of the case to argue against intellectualism.
I don’t think these arguments are going to ultimately work. But it’s clear we are a long way from Rylean regress arguments here. And that’s where I think the debate about regress arguments should end. We have a good argument against Identity Intellectualism. And we have some suggestive considerations that seem to tell against Constitutive and Causal Intellectualism, but whether these arguments ultimately work will depend on considerations independent of the regress.
Objection: Stanley and Williamson (2017) have recently defended the idea that skill is a disposition to form knowledge. And they back this up with empirical analysis of intelligent motor skills, especially drawing on the survey by Yarrow, Brown, and Krakauer (2009). Is this kind of intellectualism subject to the regress worries?
Reply: Once we are taking the dispositions themselves to be the skills, not the underlying knowledge, it feels that we are a long way from traditional intellectualism. But the view is independently interesting, and it is a useful segue to thinking about the relationship between intellectual skills, as conceived of in this paper, and motor skills.
I’ve already mentioned that the Bob example does not seem to fit well with Stanley and Williamson’s paradigm. And there is something suspicious about a theory of physical skill that divorces it so strongly from the physical. To be a skilled batsman requires more than dispositions to get knowledge, one might suspect. Stanley and Williamson have a reply to this suspicion. They write,
Consider the difference between someone who can bench-press a maximum of 100 pounds and someone who can bench press 150 pounds. We may suppose that both employ the same technique; only brute strength makes the difference between them. Both are equally skilled ...Any view of skill must account for such cases. In particular, it must explain why strength, speed, and stamina are not themselves skills.(Stanley and Williamson 2017, 9, page references to preprint)
But even if strength is not a skill, it might be a prerequisite for a skill. A batsman whose degenerative back condition means he lacks the flexibility to deploy his trademark pull shot has lost a skill, even if he hasn’t lost any dispositions to form knowledge. There is a puzzle as to why qualitative physical differences matter so much to skill attributions why quantitative ones do not. If you can’t turn to pull the ball, you’ve lost a skill, but if a muscle strength decline reduces the power of your pull shot, your skills haven’t declined. But that difference doesn’t justify making skills entirely cognitive.
Still, there is a cognitive angle. One central point of this paper is completely consistent with Stanley and Williamson’s picture; motor skills often require forming the right knowledge. The skilled batsman doesn’t just pick up many characteristics of the bowler’s delivery, they pick up the ones that are most relevant to the trajectory of the ball. As the Bob example shows, they also have to activate that knowledge for it really to be a skill, but that’s not a new objection.
There is one other cognitive aspect of motor skill that Yarrow, Brown, and Krakauer (2009) draw attention to, and which fits very nicely with the theme of this paper. It’s a specific instance of a much more wide-ranging skill. Sometimes an agent knows that in some time some evidence, drawn from a large space, will come in. She will shortly thereafter have to act in response to the evidence. She has some time to plan now. What should she do? In many such cases, backwards induction is impossible; there are too many possible pieces of evidence that could come in, and planning for each of them is a waste of resources. On the other hand, not planning at all is also a waste of the time she now has, and will lack once the evidence comes in. The solution is to do some planning. And there is a real skill involved in getting the resource allocation right, and neither wasting effort planning for unlikely scenarios, nor wasting the ability to be prepared before one needs to act.
Yarrow, Brown, and Krakauer (2009, 590–91) suggest the same thing happens at a very low level. Highly skilled athletes are making many places in advance of knowing exactly how they will act. Part of the skill involved is allocating the right resources to each of these planning activities. Many of them will ultimately be wasteful, since they are plans for eventualities that do not arise. And one failure condition is that a single plan is not selected, and the agent performs some combination of multiple plans that are worse than either one plan. That failure state is part of the evidence that there is this low-level planning going on before actions. But it is a real skill, and part of the skill is focussing on just the right things.
So motor skills often have as a constituent part intellectual skills. Some of those skills are closely tied to knowledge; for instance, having priors that track frequencies. Sometimes the skill involved is in focussing on the evidence that the posterior probability is maximally sensitive to, and reacting to that evidence. Sometimes the skill is not attending to evidence that is just going to be unhelpful noise in the activity in question (Yarrow, Brown, and Krakauer 2009, 589). And sometimes it is in allocating just the right resources to forward planning. All of these seem like intellectual skills, and parts of motor skills. We could try to squeeze all of them into a framework of being dispositions to form knowledge, but it seems more perspicuous to just present the plurality of ways in which the intellect and the body interact, rather than trying to find a single framework.
Objection: Appeal to skill does not stop the regress. If we need to posit something, say a skill, that comes between the possession of knowledge and the use of knowledge in reasoning or action, then we also to posit something that comes between the possession of a skill, and the use of that skill in reasoning or action. (Compare (Stanley 2011, 26)).
Reply: What I’m going to say here is similar to what Jeremy Fantl (2011) said in a response to an earlier version of Stanley’s argument, so I’ll be brief. Skills are dispositions. We don’t need to posit anything that comes between the disposition and its triggering. If a string is disposed to produce a middle C when struck, and it is struck, we don’t need to posit an extra intermediary between the striking and the note. Dispositions stop regresses.
But, you might insist, couldn’t the same be true of knowledge? After all, on a broadly functionalist construal of the mental, knowledge is a kind of disposition. My reply is in theory knowledge could stop such a regress, but in practice it is unlikely. An agent could be facing a problem where the possible considerations and options can be enumerated without using any particular skill, and the options are few enough that they can be each considered in turn. That is the situation an agent playing a relatively simple game might face. But it isn’t the general human condition. In practice, we face problems every moment where it requires skill to bring the right considerations to bear, at least given the processing capacities we have available.
Objection: There are semantic arguments that attributions of know how are attributions of propositional knowledge. This shows that Ryle was wrong to draw a broad distinction between know how and know that.
Reply: I’m not making any claims about either know how or about ‘know how.’ I am making some claims about skill, and those imply some claims about ‘skill.’ But I’m sympathetic to the idea that reports of know how are often reports of some kind of practical propositional knowledge. I certainly haven’t offered any arguments, nor I think any considerations in the direction of an argument, against this view.
Indeed, there are a lot of intellectualst positions that I’m not arguing against here. Anti-intellectualism is often tied up with the view that there is an important distinction between theoretical and practical fields. The arguments I’ve developed here suggest that if there is such a distinction, then proving mathematical theorems is on the ‘practical’ side. I think that’s a strange enough conclusion that it is time to change our terminology. That’s why I’ve talked about the distinction between intellectual skills and knowledge, not the distinction (if such there is) between know how and know that, or between praxis and theory.
This argument is obviously rather quick, and I doubt will persuade someone already convinced of 1. For much more extensive arguments against 1, see the Smith, Ryan and Steup articles cited in the text, plus Adams (1985).↩︎
I thought this example was purely fictional, coming from the Monty Python sketch reproduced in Dempsey (2012, 741). But Ben Wolfson pointed out to me that it’s recorded as a true story in Hadley (1903, 255).
It’s actually striking how few really good off-the-cuff quips there are in recorded history. The famous one attributed to Wilde, “I have nothing to declare but my genius,” is probably apocryphal, and in any case sounds prepared. Lists of famous come-backs and ripostes are usually crowded with written responses. Word play is hard.↩︎
Or, perhaps even more plausibly, coming up with a short list of possible openings, choosing the best, and doing what one can to figure out how to complete the response while uttering the start of it. Thanks here to Ben Wolfson.↩︎
Thanks to Bernard Kobes and John Collins for helpful discussions about the chess examples.↩︎