27

Blind Reasoning 1 ABSTRACT The paper asks under what conditions deductive reasoning transmits justification from its premises to its conclusion. It argues that both standard externalist and standard internalist accounts of this phenomenon fail. The nature of this failure is taken to indicate the way forward: basic forms of deductive reasoning must justify by being instances of blind but blameless' reasoning. Finally, the paper explores the suggestion that an inferentialist account of the logical constants can help explain how such reasoning is possible. 1. THE QUESTION I'm in the mood for some music; what, I wonder, is on offer today in Carnegie Hall? A quick check of the schedule reveals that Martha Argerich is scheduled to play on the 20th. As a result, I come to believe that: (1) If today is the 20th, then Martha Argerich is playing today in Carnegie Hall. A glance at the calendar reveals that in fact: (2) Today is the 20th. With these two beliefs in place, I move immediately to the conclusion that: (3) Martha Argerich is playing today in Carnegie Hall. And I pick up the phone. If, prior to making this modus ponens inference, I already believed (presumably with only a low level of confidence) that Martha Argerich was playing today in Carnegie Hall, then the inference looks to have strengthened whatever justification I had for that belief. If, prior to making the inference, I had no views about who was playing today in Carnegie Hall, the inference looks to have augmented my beliefs with a further justified belief. Whichever scenario obtained, how did my two premises contribute to justifying the conclusion that I drew on their basis? Under what conditions does an inference transfer justification in this way? Clearly, at the very least, the following two conditions must be satisfied. First, the thinker must be justified in believing the premises. Second, his justification for believing the premises must not depend on his being antecedently justified in believing the conclusion. Equally clearly, though, these conditions do not suffice for the inference to transfer justification. In addition, the premises must bear an appropriate relation to the conclusion they ground. And my question is: What is that relation? In this paper, I am going to restrict myself to asking this question about deductive inference, leaving it an open question to what extent what is said here generalizes to other cases of justification- or warrant-transfer.2 In a deductive inference, the thinker takes his premises to justify his conclusion in part because he takes them to necessitate it.3 II. INFERENTIAL EXTERNALISM The simplest possible answer to our question is this: (Simple Inferential Externalism): A deductive inference performed by S is warrant-transferring just in case (a) S is justified in believing its premises (b) S's justification for believing its premises is suitably independent of his justification for believing the conclusion, and (c) the implicated pattern of inference is valid-necessarily such as to move S from truth to truths.4 This can't be a good answer: large numbers of inferences that we are in no intuitive way justified in performing satisfy the stipulated conditions. For example: it is easy for me to be justified in believing any particular claim of the form: x, y, z, and n are whole numbers and n is greater than 2. If I inferred from this proposition that: x° + y° is not equal to z' I would have performed an inference that is, as we now know, reliably truthpreserving. But it would be absurd to suppose that anyone making such an inference would be drawing a justified conclusion, whether or not they knew anything about Andrew Wiles's proof of Fermat's last theorem or had checked each individual inequality. Being valid (in conjunction with the other two conditions) clearly does not suffice for being warrant-transferring, even if it may be necessary.5 A number of philosophers seem to believe that this objection can be easily met. All we need to do, they say, is to restrict the Reliabilist claim to those inferences that are sufficiently 'simple'.6 But what couldsimple’ possibly mean here if not something like: an inference whose validity it is easy to take in'? If that isn't whatsimple’ means, why should an inference’s simplicity be relevant to the question of justification? If that is what simple' means, then it's this presumedtaking in’ that’s doing the relevant explanatory work, and not the assumed reliability of the simple inference. In a moment, we will consider views according to which an inference is entitling just in case the thinker can take in' its validity, but this is not where such views belong. The counterexamples to Simple Inferential Externalism echo a broader class of objections to Reliabilist accounts of justification more generally. Since many of these examples are so well known, I will not discuss them here in detail except to say this. I am not so impressed with those counterexamples that rely on the subject's having a justified belief to the effect that some reliable beliefforming method of his is not reliable; I think a Reliabilist can handle those cases by imposing ano-undermining belief’ condition. But I am inclined to regard as decisive those counterexamples in which the reliability of the relevant method is not subjectively undermined-either because the subject has no justified belief about its reliability, or because he has no belief about it at all. Laurence Bonjour has described just such a case. Norman, under certain conditions which usually obtain, is a completely reliable clairvoyant with respect to certain kinds of subject matter. He possesses no evidence or reasons of any kind for or against the general possibility of such a cognitive power or for or against the thesis that he possesses it. One day Norman comes to believe that the President is in New York City, though he has no evidence either for or against this belief. In fact the belief is true and results from his clairvoyant power under circumstances in which it is completely reliable.7 Our robust response to this case is that Norman is not justified. And a plausible and widelyaccepted diagnosis of our response is that we are reluctant to regard someone as justified in holding a given belief if they are being epistemically irresponsible in holding that belief. Being justified is, at least in part, a matter of being epistemically blameless. What lesson should we draw from these counterexamples? The moral seems almost forced. We need to ensure that being justified excludes being epistemically blameworthy. Mere reliability doesn’t do that. How is that to be ensured except by insisting that, if a subject is to be justified in believing some proposition p, he must have to hand a reflectively accessible warrant for the proposition that p? It looks, in other words, as though the counterexamples to Reliabilism motivate an Access Internalism about justification: S is justified in having the belief that p only if S is in a position to know,8 by reflection alone, that he has a warrant for the belief that p. If S is to have genuine justification, it must be a reflectively transparent justification.9 III. INFERENTIAL INTERNALISM If we go along with this diagnosis of the failure of Reliabilism and apply it to the case we are focusing on-the case of warrant transfer by deductive inference-the thought would then be that, in a deductive inference, the subject must be in a position to know by reflection alone that his premises provide him with a good reason for believing the conclusion, if his inference is to justify his conclusion. Call this (Simple Inferential Internalism): A deductive inference performed by S is warrant-transferring just in case (a) S is justified in believing its premises (b) S’s justification for believing its premises is suitably independent of his justification for believing the conclusion, and (c) S is able to know by reflection alone that his premises provide him with a good reason for believing the conclusion. There are any number of problems with this idea. Take the case of our original Argerich modus ponens (MPP) inference, from (1) and (2) to (3). What would it be for S to be able to know by reflection alone that his premises provide him with a good reason for believing the conclusion? Since we are dealing with a deductive inference, the most natural suggestion is that S would know this epistemic fact only if he knew that his premises necessitate his conclusion. So the question becomes: How might S be in a position to know by reflection alone that p and p q' imply q? There look to be two options: inferential and non-inferential. The inferential route, which would include empirical and broadly pragmatic accounts of knowledge of logical implication, may be ruled out immediately. The point is a subtle one. It's not that there is necessarily a problem with rulecircular justifications of meta-logical claims. Indeed, I have elsewhere argued that, if we are to have knowledge of basic logical or meta-logical truths at all, it must be via rule-circular reasoning.1° However, in the present context, where knowledge of the validity of MPP is supposed to be a component in the justification that we have for reasoning according to MPP, an inferential route to knowledge of the validity would be completely useless. The very sort of reasoning whose justification is at issue would have been presupposed. What about the non-inferential options? If a given item of knowledge is noninferential, then it is either justified by observation alone or it is justified by nothing. For reasons that I don't have the space to rehearse here, it seems to me very implausible that one can be said to know the general proposition that any argument of the form MPP is valid on the basis of nothing, as though all one would have to do to be justified in believing such an ambitious proposition is simply to believe it.l1 If there is to be any hope for the non-inferential option, it must lie along the observational branch. IV. RATIONAL INSIGHT Clearly, what is at issue is not ordinary empirical observation. Rather, the idea is that we are equipped with a special capacity for non-empirical observation, a capacity whose exercise is capable of yielding insights into necessary truths. That we have such a capacity has recently been defended by Laurence Bonjour. When I carefully and reflectively consider the ... inference ... in question, I am able simply to see or grasp or apprehend ... that the conclusion of the inference must be true if the premises are true. Such a rational insight, as I have chosen to call it, does not seem to depend on any particular sort of criterion or any further discursive or ratiocinative process, but is instead direct and immediate.12 In finding the idea of rational insight attractive, Bonjour joins a venerable tradition, one that stretches from Plato through Leibniz to Godel. Why, then, does the idea have so few supporters these days? The single most influential consideration against rational insight theories can be quite simply stated: no one has been able to explain, clearly enough, in what an act of rational insight could intelligibly consist. No one denies, of course, that we can think about properties and relations, including logical properties and relations, and that, as a result, we can reason our way to general conclusions about them, just as we can reason our way to conclusions about other topics. The question is whether we can be said to have some sort of non-discursive, non-ratiocinative, insight into their natures, an insight that would disclose immediately, and without the help of any reasoning whatsoever, that all instances of MPP are truth-preserving. Bonjour does not succeed in convincing us that there is. On reading his book, one can't help but be struck by how little progress a resolute and resourceful defender of rational insight was able to make. There are many issues that could be discussed with profit; here I have space only to raise the fundamental issue of cognitive access. What sort of relation obtains between a thinker and a property, when the thinker hasrational insight’ into its nature? Or, to pick a more immediately relevant example, what relation obtains between a thinker and the conditional, when the thinker has a rational insight into its nature, when he is simply able to see that when his MPP premises are true so must be its conclusion? The analogy with sense perception encourages us to think that the relation between the thinker and the conditional is causal. But that is impossible since the conditional is just an abstract object. Bonjour wrestles with this issue and comes up with a three-part response. The first is to downplay the analogy with sense perception. The second is to propose the revival of an Aristotelian conception of mental content, according to which when we think about a given property -triangularity, for example-that very property itself is instantiated in our thought. Finally, Bonjour suggests assimilating rational insight to a form of introspective examination of that thought content. His proposal raises more questions than it answers. First, we are not really told how the property of triangularity could be instantiated by my thoughts without their being-absurdly-triangular as a result. Second, we are not told how this neo-Aristotelian conception of mental content would help, even if it were not mystifying. Suppose my thoughts about a particular triangle do instantiate the property of triangularity. How does this help explain how I am able to directly divine the nature of triangularity without the help of any reasoning whatsoever? The belief that the sum of the interior angles of any triangle is 180 degrees and the belief that it is 320 degrees both instantiate triangles. Why is the first justified and the second not? We are left staring at the problem with which we began, rather than feeling that we have been placed on the path to real understanding.