29

VII. BLIND YET BLAMELESS INFERENCE: CONCEPT CONSTITUTION An important question-which I don’t wish to prejudge for present purposes-is whether the validity of the inference is a necessary condition on warrant transfer. What I will be exclusively concerned with in the remainder of this paper is the question of blamelessness. The inflationary' answer to that question that I want to explore may be roughly formulated as follows (we'll see how to refine it later): A deductive pattern of inference P may be blamelessly employed, without any reflective appreciation of its epistemic status, just in case inferring according to P is a precondition for having one of the concepts ingredient in it. Now, in some sense this is a very old answer to our question. It falls into what may broadly be calledanalytic’ explanations of the a priori. Previous versions of such views, however, have suffered, I believe, in two respects .21 First, they did not adequately distinguish between questions concerning our entitlement to certain logical beliefs, and questions concerning our entitlement to certain belief-forming methods of inference (like the one I have dubbed MPP). Second, they did not adequately explain what concept-constitution has to do with the epistemology of blind inference. In both of these respects, I hope here to do a little better. Prima facie, there is a difficulty seeing how appeal to concept constitution can help with our question. What is the connection supposed to be? The thought is this. Suppose it’s true that my taking A to be a warrant for believing B is constitutive of my being able to have B-thoughts (or A-thoughts, or both, it doesn’t matter) in the first place. Then doesn’t it follow that I could not have been epistemically blameworthy in taking A to be a reason for believing B, even in the absence of any reason for taking A to be a reason for believing B? For how could I have had antecedent information to the effect that A is a good reason for believing B, if I could not so much have had a B-thought without taking A to be a reason for believing B in the first place? If inferring from A to B is required, if I am to be able to think the ingredient propositions, then it looks as though so inferring cannot be held against me, even if the inference is blind. Applied to the case of deductive inference before us, then, the thought would be that we would have an explanation for the blameless blindness of MPP if it’s constitutive of having the concept conditional that one take p and p q' as a reason for believing q. Now, of course, the idea that, in general, we come to grasp the logical constants by being disposed to engage in some inferences involving them and not in others, is an independently compelling idea. And the thought that, in particular, we grasp the conditional just in case we are disposed to infer according to MPP is an independently compelling thought. So, if the meaningentitlement connection that I've gestured at is correct, it looks as though we are in a position to mount an explanation of the blameless blindness of MPP that we were after. VIII. PROBLEMS FOR THE MEANING-ENTITLEMENT CONNECTION Unfortunately, matters are not quite so straightforward. If we spell out the principle underlying the meaning-based explanation of blameless blindness gestured at, it would be this: (Meaning-Entitlement Connection, or MEC): Any inferential transitions built into the possession conditions for a concept are eo ipso entitling. And the trouble is that, at least as stated, there seem to be clearcut counterexamples to the MEC: it doesn't in general seem true that if my taking A as a reason for believing B is constitutive of my believing B, that this automatically absolves me of any charge of epistemic blameworthiness. For there seem to be clear cases where the acceptance of some inference is written into the possession of a given concept but where it is also clear that the inference isn't one to which the thinker is entitled. One famous illustrative case is Arthur Prior's connectivetonk’.22 To possess this concept, Prior stipulated, a thinker must be willing to infer according to the following introduction and elimination rules: (Tonk) A A tonk B A tonk B B Obviously, no one could be entitled to infer any B from any A; but this entitlement appears to flow from the possession conditions for tonk' along with the MEC. A similar conclusion can be drawn from the case of racist or abusive concepts, for example the concept boche discussed by Dummett.23 Plausibly, a thinker possesses the concept boche just in case he is willing to infer according to the following rules: (Boche) x is German x is boche x is boche x is cruel Yet no one is entitled-let alone simply as the result of the introduction of a concept into the language-to the view that all Germans are cruel. How should we think about such cases? Robert Brandom has this to say about boche-like concepts: The use of any concept or expression involves commitment to an inference from its grounds to its consequences of application. Critical thinkers, or merely fastidious ones, must examine their idioms to be sure that they are prepared to endorse and so defend the appropriateness of the material inferential commitments implicit in the concepts they employ ... The proper question to ask in evaluating the introduction and evolution of a concept is not whether the inference embodied is one that is already endorsed, so that no new content is really involved, but rather whether the inference is one that ought to be endorsed. The problem withboche’ is not that once we explicitly confront the material inferential commitment that gives the term its content it turns out to be novel, but that it can then be seen to be indefensible and inappropriate-a commitment we cannot become entitled to.24 From the standpoint of a proponent of the MEC, there is nothing in this passage that helps protect it from the threatening examples. It’s no answer to the challenge they pose to observe that whatever entitlement concept possession gives rise to, it can be defeated by further considerations. No one should expect more than a defeasible entitlement, even from concept possession; and what’s implausible in the case of tonk' andboche’ is that there is any entitlement there at all, defeasible or no. If we are to save the MEC, it seems to me that we must do one or both of two things: either restrict it to certain concepts from which entitlement really does flow, or restrict what we count as a genuine concept. I will advocate doing the former. The latter strategy is suggested by the work of Christopher Peacocke who has long urged that we should require that the meaning-constituting rules of a genuine concept be truth-preserving.25 If we adopt this requirement, we can say that what’s wrong with both tonk' andboche’ is precisely that there is no concept that those terms express, for there is no reference for tonk' andboche’ that’s capable of making all of their constitutive rules truth-preserving. While this might seem to yield the right result for tonk' it doesn't yield the right result forboche’: it’s hard to believe that racists who employ boche-like concepts fail to express complete thoughts. And even if we were to put this complaint to one side, it seems clear that truth-preservation alone will not suffice for dealing with our problem about the MEC. Imagine someone theorizing about water and coming to believe, for whatever reason, that the way in which it is correct to say that water is composed of H2O is that there is some other stuff that composes water and that it is composed of H2 0.26 So he introduces a term-aqua'-to name this stuff and he stipulates that it is to be governed by the following introduction and elimination rules: (Aqua) x is water x is aqua x is aqua x is H2O Unlike the case ofboche’, we have no independent reason for thinking that these rules are not truth-preserving. But there is clearly something fishy about this concept. And however one feels about that, there is certainly a problem for the MEC, given only the resources that we’ve been accorded so far: for no one could think that the mere act of introducing the concept aqua into one’s repertoire could give one a priori entitlement to the inference from x’s being water to x’s being H2O. Or consider the conceptflurg individuated by the following introduction and elimination rules: It turns out to be a result that Wiles had to prove on the way to proving Fermat’s Last Theorem that every elliptical equation can be correlated with a modular form (the Taniyama-Shimura conjecture). Once again, therefore, we have no independent reason to think that these introduction and elimination rules are not necessarily truth-preserving. But it’s hard to see that one is a priori entitled, merely on the basis of introducing the term flurg', to the Taniyama-Shimura conjecture. So there is still a problem for the claim that entitlement flows from meaning-constitution, given only the requirement that a concept's introduction and elimination rules be truth-preserving. IX. DEFECTIVE CONCEPTS AND BLAMELESS INFERENCE I would like to propose a different diagnosis of what has gone wrong with concepts such as aqua and flurg, one that doesn't depend on denying that they constitute genuine thinkable contents. That denial can be sustained, I believe, in the case oftonk’, but that is a rather extreme case. No one can actually possess the concept allegedly expressed by tonk', because it isn't possible for someone to follow the rules that are constitutive of that connective. For to follow those rules one would have to be prepared to infer anything from everything, and that is no longer recognizable as belief or inference. But no such extreme claim can be made with respect to all the other examples that have been causing problems for the hypothesized connection between meaning and entitlement. Start with the example ofaqua’. The theorist who has conceived the need to introduce a term for the concept aqua has come to hold the following theory: There is some stuff, distinct from water, that composes water and that is itself composed of H2O. Let me call it "aqua".' Such a theorist already believes in water and H2O, we may suppose. He has come to hold an additional belief about the world, namely, that it contains another substance, one that is related to water and H2O in the specified way. Now, the way we have written down the inferential rules foraqua’ essentially amounts to insisting that, in order to have the concept aqua you must be prepared to believe this little aqua theory. Given that you already believe in water and H2O, the only way for you to acquire the concept aqua, on this account of its inferential rules, requires you to believe that there is such a thing as aqua. One cannot so much as have the concept of aqua without being prepared to believe that there is such a thing. And although it seems to me that one can define and then think in terms of such a concept, it does seem like an epistemically questionable thing to do. Even if the aqua theorist were certain that there is such a thing as aqua, he should want the concept he expresses by that term to leave it open whether there is. He should allow for the conceptual possibility that he is mistaken; and he should certainly allow others intelligibly to disagree with him about aqua’s existence. The concept itself should not be designed in such a way that, only those who believe a certain creed are allowed to possess it. Ordinary scientific terms in good standing-`neutrino’ for example-are held to have just this feature, of intelligibly allowing for disagreement about their extension. Thus, we don’t think of the rules which correspond to our possession of the concept neutrino as consisting in the propositions that would actually be believed by a proponent of neutrino theory, but rather as corresponding only to what someone would be willing to believe who was conditionalizing on the truth of neutrino theory. If we follow Russell, Ramsey, Carnap and Lewis, and represent neutrino theory T(neutrino) as the conjunction of the two propositions (S) (ax) Tx and (M) (ax) Tx – T(neutrino) then the point is that we think of possession of the concept neutrino as requiring someone to affirm only M and not S as well.27 Now, someone could certainly introduce a concept that did not have the conditionalized structure that I’ve claimed is actually true of neutrino, but which consists rather in the inferences that are characteristic of neutrino theory unconditionalized. Call this neutrino+. Such a person would insist that it is a condition on having his concept of neutrino that one be willing to endorse the characteristic claims and inferences of neutrino theory, and not merely the conditionalized claim captured in (M). But, for the reasons previously articulated, there would be something epistemically defective about this concept, even if its constitutive rules turned out to be truthpreserving. Flurg, aqua and neutrino+, then, all suffer from the same problem: they are all unconditional versions of a concept, when only its conditionalized version would be epistemically acceptable. I don’t think we should put this by saying that they are not real concepts. Concepts are relatively cheap. But they are defective concepts. They are structured in such a way that perfectly reasonable questions about their extensions are foreclosed. Under what conditions is only a conditionalized version of a concept acceptable? Here I want to make two claims: one bold, one sober. (Bold) Whenever both a conditional and an unconditional version of a given concept are available, it is the conditional version that ought to be used. Given the availability of both versions, the unconditional version counts as epistemically defective. (Sober) In the case of some concepts, only the unconditionalized version will be available. Start with Bold. Whenever a conditionalized version of a concept is available, that is the version that ought to be used. In those contexts, an unconditionalized version would be defective. The argument for this is quite straightforward and recapitulates the considerations we have just been looking at. You don’t ever want the possession conditions for a concept to foreclose on the possible falsity of some particular set of claims about the world, if you can possibly avoid it. You want the possessor of the concept to be able coherently to ask whether there is anything that falls under it, and you want people to be able to disagree about whether there is. If in a certain range of cases, however, it is logically impossible to hold the governing theory at arm’s length then, in those cases, obviously, it can hardly be a requirement that one do so. But in all those cases where that is possible, it ought to be done.