麦肯锡 困境McKinseyParadox

1 Published in Synthese 130 (2002), 279-302. MCKINSEY PARADOXES, RADICAL SCEPTICISM, AND THE TRANSMISSION OF KNOWLEDGE ACROSS KNOWN ENTAILMENTS ABSTRACT. A great deal of discussion in the recent literature has been devoted to the so-called ‘McKinsey’ paradox which purports to show that semantic externalism is incompatible with the sort of authoritative knowledge that we take ourselves to have of our own thought contents. In this paper I examine one influential epistemological response to this paradox which is due to Crispin Wright and Martin Davies. I argue that it fails to meet the challenge posed by McKinsey but that, if it is set within an externalist epistemology, it may have application to a related paradox that concerns the problem of radical scepticism. 1. THE MCKINSEY PARADOX According to semantic externalism, some of a subject’s thought contents are individuated, at least in part, by factors within that subject’s environment.1 Seemingly, however, such a theory stands in direct opposition to the highly intuitive thesiswhat I shall call the ‘privileged access’ thesisthat one can have a priori knowledge of one’s own thought contents, where this is understood as the claim that one can know them purely by reflection (i.e., without an empirical investigation). Crudely put, the concern is this: If semantic externalism and the privileged access thesis are both true, then it would seem to follow that one could gain knowledge of one’s ‘wide’ thought contentsand thus of non-mental empirical factsby conducting a purely a priori investigation.2 Much of the recent debate in this respect has tended to cluster around a neat formulation of the tension due to Michael McKinsey (1991) which we shall represent as follows, where ‘KAP’ refers to a priori knowledge: The McKinsey Argument (1) KAP [(MK1) I have a mental property, M]. (2) KAP [(MK2) If I have mental property M, then I meet external condition E]. (3) KAP [(MK3) Therefore, I meet external condition E]. The first line of the argument simply captures a paradigm case of an agent’s a priori knowledge, gained via privileged access, that he has a certain mental property. The second line concerns that same agent’s a priori knowledge, via the philosophical thesis of semantic 2 externalism, that the possession of this mental property entails that a certain external condition has been metan empirical non-mental truth about that subject’s environment. Finally, the third line puts these two claims together to conclude that the subject must, therefore, have a priori knowledge that this empirical condition has obtained. Intuitively, since the external condition in question seems precisely the sort of thing that cannot be known a priori, this represents a reductio of the supposed compatibility of semantic externalism and privileged access. One of them must go. In order to see this incompatibility in more detail, consider the following influential argumentdue to Hilary Putnam (1975)which has been used to motivate a form of semantic externalism. Let twin earth be like earth in every respect except that there is no water, just a qualitatively indistinguishable substance‘twater’that has the abbreviated microstructure XYZ. Now suppose that I, and my twin on twin earth, both have exactly the same narrow mental states,3 and are equally as ignorant of chemistry. According to Putnam, even despite our shared narrow mental states, our respective utterances of sentences containing the word ‘water’ will have different semantic properties. On earth, my use of this word will refer to H2O, whereas on twin earth, my twin’s use of this word will refer to XYZ. On the assumption that meaning determines reference (so that a difference in reference suffices for a difference in meaning), it follows that my use of this word will mean water while my twin’s use of the word will mean twater. Hence, as Putnam famously put it, “meanings ain’t in the head”. According to Paul Boghossian (1997), however, the combination of this proposal and privileged access leads to a priori knowledge of empirical truths. Boghossian takes Putnam to be arguing that the possession of water thoughts entails the existence of water (of H2O). As a result, it would appear to follow that a priori knowledge of one’s thought contents in this instance, coupled with a priori knowledge of the truth of semantic externalism (and its application to this particular concept), would entail a priori knowledge of the fact that water exists, which is clearly an empirical truth about the world. We thus get the following McKinsey-style argument: The Putnam-McKinsey Argument (1) KAP [(P1) I have the concept water]. (2) KAP [(P2) If I have the concept water, then water exists]. (3) KAP [(P3) Water exists]. And since the existence of water seems to be the very sort of thing that is settled by empirical investigation and not by armchair reflection, it follows that we have an instance of the 3 McKinsey style reductio of the conjunction of privileged access and semantic externalism that we saw above. One natural way to respond to a paradox of this sort is to decline to take the puzzle at face value and instead undertake a detailed investigation into the nature of the theses of semantic externalism and privileged access in order to see whether the argument can be blocked or, despite the odds, the conclusion accepted. In contrast, the more unorthodox approach to the problem that I shall be considering in this paper involves granting the truth of something like the above formulations of these two theses and investigating instead the epistemological intuitions that are at work here. It is important to note that this approach runs orthogonal to the standard line because, in principle at least, they could both offer plausible solutions to this puzzle. It could be, for instance, that not all construals of semantic externalism and privileged access license this problematic conclusion and that certain epistemological considerations show that, in any case, the inference to a priori knowledge does not go through. Accordingly, the epistemological account of the problem need not necessarily be viewed as being in competition with any of the current non-epistemological responses on offer. 2. THE WRIGHT-DAVIES PROPOSAL Of the epistemological attempts to dissolve the McKinsey paradox, the most influential has been that put forward by Martin Davies (1998) and developed by Crispin Wright (2000a; cf. Wright 1997). Essentially, they argue that what arguments of this form expose is the fact that even valid arguments can have premises that are true and known and yet the conclusion be such that this knowledge will not transmit across the entailment to the conclusion.4 The best way to understand this notion of ‘Transmission’ is in contrast to the related (and weaker) epistemic principle of ‘Closure’ that Davies and Wright both accept. This can be (roughly) expressed as follows: KC: {K [ϕ] & K [ϕ → ψ]} → K [ψ] In words, if an agent knows a proposition, ϕ, and knows that ϕ entails a second proposition, ψ, then that agent knows ψ also.5 More informally still, one might express this principle as the claim that knowledge always transfers across known entailments. Note that the McKinsey template that we have just considered pivots upon an instance of this principle (where the ‘K’ operator picks out a priori knowledge throughout). Denying Closure could thus be one way in 4 which one might respond to this puzzle.6 This is not, however, the course of action taken by Davies and Wright because they believe that there is a stronger sister principle to Closure in play here, which they call ‘Transmission’, and that it is this principle that is essential to the formulation of the paradox. Transmission is held to be a stronger principle than Closure because it makes the extra demand that the knowledge that transfers across the known entailment should also preserve what Wright refers to as the “cogency” of the argument, which is its aptitude to produce rational conviction. Here is Wright: A cogent argument is one whereby someone could be moved to rational conviction of the truth of its conclusion. (Wright 2000a, 140) In essence, the idea is that the McKinsey argument, though not a counterexample to Closure, is nevertheless a case in which Transmission fails. That is, although the knowledge at issue in the premises will transfer to the conclusion (in the sense that if the premises are known then the conclusion must be known also), that knowledge will not transmit to the conclusion because the argument lacks cogency. We shall examine the reasons that Davies and Wright offer for this contention in a moment. First, however, it is worthwhile outlining what dialectical consequences it has. To begin with, it is important to note that since Davies and Wright retain Closure it follows that if the agent does know the premises ((MK1) and (MK2)) of the McKinsey argument, then he must also know the conclusion, (MK3). What the Davis-Wright line denies is thus not that the agent might know the conclusion, but rather the more problematic contention that the agent knows the conclusion on the basis of legitimately inferring that conclusion from his knowledge of the premises. This point is important because what is problematic about the argument is not that the agent knows the conclusion, but rather that this knowledge has been derived in a purely a priori fashion by inferring it from the premises. If the Davis-Wright manœuvre holds water, then this latter claim will be blocked because the agent will be unable to legitimately infer knowledge of the conclusion from his knowledge of the premises, and thus the argument will offer us no reason for thinking that the agent knows the conclusion on a priori grounds. According to Davies and Wright, the reason why the McKinsey argument lacks cogency is because it is question-begging in the sense that an epistemic entitlement to believe the conclusion is already presupposed, without independent motivation, in the epistemic entitlement that the agent takes himself to have for believing the premises. In particular, what the McKinsey argument illustrates, according to Davies and Wright, is that an epistemic 5 entitlement to believe that certain external conditions have obtained is already taken for granted in one’s putative a priori knowledge of the wide contents of one’s own mental states. Given that this is so, it should come as no surprise to find that Closure holds since if one has knowledge of the premises and such knowledge presupposes an epistemic entitlement to believe the (true) conclusion, then how could it not be that the conclusion is not also known? Nevertheless, such an argument would provide no reason whatsoever to believe the conclusion of the argument, and thus Transmission would fail. One who doubted the conclusion would, for instance, be offered no reason to change his mind upon being presented with such a question- begging argument. As Wright expresses the matter: Here is the point in a small nutshell. Suppose that I want to describe myselftruly as it happensby using the words: “I believe that water is wet”. If some form of externalism is true which would be strong enough to sustain the second premise of the McKinsey argument, then there are external preconditions of my expressing a true belief by those words whose satisfaction I may nevertheless, without compromise of the warrant for my claim, have done nothing special to ensure. Can that warrant now licitly be extended to the claim that those preconditions are met? It should seem obvious that it cannot, for the simple reason that the warrant is in the first place conditional on the concession, as it were, that unless there is extant positive reason to doubt that the external conditions are met, the possibility that they are not can be ignored. Given that the warrant I start out with has this concession- dependent character, it naturally cannot be massaged by inference into a reason for a positive view about the issue which the concession was precisely a concession to take for granted. (Wright 2000a, 156) In a still smaller nutshell, we might put the point by saying that one only has a priori access to one’s wide thought contents on the assumption that certain empirical conditions are legitimately disregarded from the outset such as the possibility that, appearances notwithstanding, one in fact lives on a ‘dry’ earth on which there is no water. Accordingly, one cannot, contra the McKinsey argument, reason one’s way to a priori knowledge that such presuppositional empirical conditions obtain. Of course, unlike other arguments where it is clear that the question is being begged in this way, it is far from transparent that the McKinsey argument is question-begging. The guiding diagnostic thought that underpins the Davies-Wright proposal, however, is that in every discourse there are propositions which play a presuppositional role(and hence which one cannot know via participation in that discourse)but that sometimes these presuppositions are not recognised as such by the participants of that discourse. By making explicit which propositions perform this ‘presuppositional’ function in any given discourse we can thus evade a great deal of philosophical confusion. As Davies puts it: The intuitive idea [...] is something like this. In any given epistemic project, some propositions will have a presuppositional status. Suppose that the focus of project P is the proposition A, [... and] suppose that B is some proposition that has this presuppositional status in project P. Then P cannot itself yield 6 knowledge that B; nor can P play an essential role in yielding knowledge that B. (Davies 1998, 354) Or, as Wright more succinctly expresses the matter: Every epistemic project incorporates certain presuppositions which, whatever its upshot, it cannot provide warrant for. (Wright 2000a, 150)7 Note, however, that to say that knowledge of a presupposition of a certain epistemic project cannot be gained via that project is not to say that the presupposition is unknowable, only that it cannot be known via that route. This point is important because the truth of the presupposition will no doubt be entailed by many of the truths believed in that epistemic project. Insofar as there is knowledge in that discourse (and insofar as Closure holds), then one would expect the (known) presuppositions to be known as well. The point is thus not that the (known) presuppositions of a certain epistemic project cannot be known (which would almost certainly constitute a counterexample to Closure), but rather that they are not known via the epistemic project for which they play the presuppositional role. In particular, in the specific case of the McKinsey paradox they are not known via that particular a priori argument. Davies and Wright attempt to make this line of argument compelling by showing how it serves to explain the inadequacies of other arguments that are meant to be analogous, in the relevant respects, to the McKinsey argument. One example that they both consider is G. E. Moore’s (1925; 1939) infamous ‘proof’ of an external world on the basis of his putative knowledge that he has two hands,8 an argument, which, absent the stipulation that the knowledge that is at issue always be a priori, can be expressed in terms of the familiar McKinsey template: The Moorean Argument (1) K [(M1) I have two hands]. (2) KAP [(M2) If I have two hands, then there exists an external world]. (3) K [(M3) There exists an external world]. One who already doubted the existence of an external world would not be moved to change his mind by an argument of this form. Rather, he would simply retort that in claiming knowledge of (M1) Moore is thereby already presupposing that he has an epistemic entitlement for his belief in the conclusion of the argument, since it is only on this presupposition that he is in a position to coherently claim knowledge of (M1) in the first place. That is, the sceptic we have in mind here will argue as follows. Given that there is no a priori defence of his belief in (M3),9 Moore’s knowledge of (M1) could only be supported by empirical evidence, evidence which, moreover, could only supply sufficient epistemic support for (M1) on the supposition 7 that Moore was already epistemically entitled to believe (M3). Without this assurance, the evidence would be necessarily insufficient, and so not apt to support the kind of epistemic support for belief in (M1) necessary to buttress a knowledge claim. Hence, Moore’s claim to know (M1) can only provide, at best, question-begging support for his claim to know the conclusion of the argument, (M3). Moreover, as before, note that, at least as it stands, this feature of the Moorean argument does not suffice to indicate that it is a case in which Closure fails. Intuitively, if Moore does know that he has two hands, then he must know that there is an external world. Nevertheless, the point is that such an argument to this conclusion lacks cogency. If Moore does know this conclusion, then it is not as a result of a legitimate inference from his putative knowledge of (M1) and (M2) because such knowledge already presupposes an epistemic entitlement to believe the conclusion. The Moorean argument is thus a case in which, on closer analysis, we discover that Moore is simply presupposing that which he sets out to establish. No wonder, then, that the Moorean argument can seem to completely by-pass the sceptical issue. Davies and Wright thus support their diagnosis of the McKinsey argument by showing how such an argument has an analogous form to another argument, the Moorean argument, which is more obviously question-begging and so lacking in cogency. Further support for this analogy can be offered by considering how one could just as well extend the Putnam version of the McKinsey paradox that we saw above along similar anti-sceptical lines. Take the ‘brain-in-a-vat*’ sceptical hypothesis to mean that hypothesis discussed by Putnam (1981, chapter 1) which stipulates that everyone has always existed as brains in vats coupled with the added flourish that such vats will not countenance (on pain of death for the inhabitant) the presence of H2O. The motivation behind these additions to the standard brain- in-a-vat hypothesis is that they ensure that no-one has ever had causal interaction either with water itself, or with someone else who has causally interacted with water. In this way, we can guarantee that, at least by the lights of Putnam’s account at any rate, the possession of the concept of water actually entails the falsity of the brain-in-a-vat* hypothesis. As a result, if an agent has a priori knowledge both that he possesses such a concept and of the truth of Putnam’s theory, he can therefore infer, on a priori grounds, that he knows that he is not a brain-in-a-vat*. We can thus extend the Putnam-McKinsey argument above as follows: The Extended Putnam-McKinsey Argument (1) KAP [(P1) I have the concept water]. (2) KAP [(P2) If I have the concept water, then water exists]. 8 (3) KAP [(P3) Water exists]. (4) KAP [(P4) If I have the concept water, then the brain-in-a-vat* sceptical hypothesis is false]. (5) KAP [(P5) The brain-in-a-vat* sceptical hypothesis is false]. If one regards the gaining of a priori knowledge of (P3) as paradoxical, then one should likewise have reservations about gaining a priori knowledge of (P5) in this way. Indeed, Wright himself draws a similar anti-sceptical conclusion from the Putnam argument, arguing that it would follow from the Putnam account that one could know a priori that either oneself or members of one’s speech community had had encounters wit