Hume's Skepticism as a Path to Knowledge — Epoche B2
Custom Without Warrant: Hume, Circularity and Reliabilism An ancestry that runs the wrong way It has become common to present David Hume as an early externalist about knowledge [1] . The story runs like this: having shown that reason cannot justify our causal inferences, Hume turns to describe the mechanism that produces them — custom — and thereby moves epistemology from the search for reasons to the study of belief-forming processes [2] . That is the shape of modern reliabilism, so Hume is said to anticipate it [3] . This essay argues that the ancestry runs the other way. Hume is not the forerunner of reliabilism; he is the source of the strongest standing objection to it. His argument in A Treatise of Human Nature (1739–40) is that no experience can establish that experience is a guide, because any such establishment would itself be a causal inference and so would presuppose what it was meant to secure. That is the circularity that modern epistemologists rediscovered as the bootstrapping problem, and it is a problem for reliabilism specifically. A second claim follows, and it is the more useful one. The naturalist reading of Hume, which goes back to Kemp Smith's study of 1941 and is right about what it asserts [4] , has been taken to imply more than it does. Hume does have a normative epistemology, but it is not a criterion of justification. It is a set of rules for conducting causal inference well, given that we are going to conduct it — internal to the practice, silent about whether the practice is validated. Both claims can be denied, and the strongest denial is set out below. The argument, and where it occurs Hume divides the objects of enquiry into relations of ideas and matters of fact. The first are established by thought alone and cannot be denied without contradiction. The second concern what exists, and their contraries are always conceivable: that the sun will not rise tomorrow is a perfectly intelligible supposition, as he puts it in the Enquiry concerning Human Understanding (Section IV, Part I). Our beliefs about matters of fact not presently observed rest on causal inference, and the argument about causal inference is at Treatise 1.3.6. Hume asks what carries the mind from an observed impression to the belief in something unobserved. Not demonstration: no contradiction attends the denial. Not experience by itself: experience delivers past conjunctions, and to get from those to a claim about the unobserved one needs the supposition that the course of nature continues uniformly. And that supposition cannot be established by experience, since any argument for it would have to move from observed cases to unobserved ones, which is the very step at issue. Hume's conclusion is stated as a claim about a faculty: the inference is not determined by reason. Custom does it. Having seen flame conjoined with heat many times, the mind passes from one to the other without any intervening step of reasoning, and the passage carries the feeling of belief with it. What custom is not It is worth being exact about the status of this account, because the proto-reliabilist reading depends on blurring it. Reliabilism, in the form given by Alvin Goldman in 'What Is Justified Belief [5] ?' (1979), is a normative theory. A belief is justified when it is produced by a process type that in fact tends to produce truths. The theory is externalist: the believer need not be able to show, or even to know, that the process is reliable. The truth-ratio does the justifying whether or not anyone is aware of it. Now ask what would make custom reliable on Hume's account. Custom is individuated by frequency of past conjunction: it fires when two kinds of impression have been repeatedly paired. Whether a process so individuated tends to produce truths depends entirely on whether nature continues uniformly. If it does, custom is highly reliable; if it does not, custom is a machine for error, and it would go on operating in exactly the same way. Hume's point about the uniformity supposition is precisely that nothing available to us settles which case we are in. So the fact reliabilism needs is the one Hume argued is beyond reach. That does not by itself refute reliabilism — as we shall see, the externalist has a reply — but it does mean that calling Hume a proto-reliabilist misdescribes the relation. He identified the mechanism, and he identified the gap between the mechanism and any truth-connection, and he did not close the gap. Circularity, then and now The structural point in Hume's argument is that a faculty cannot certify itself using its own outputs. Modern epistemology met the same point again from the other end. William Alston's work on epistemic circularity puts it in terms of sense perception [6] : any argument that perception is reliable will have perceptual premises. Jonathan Vogel's 'Reliabilism Leveled' (2000 [7] ) sharpens it into the bootstrapping objection. If a process is reliable, then on reliabilism its outputs are justified; one may then compile those outputs, note that they have all been correct, and arrive at the justified belief that the process is reliable — without ever having had independent evidence of it. Reliabilism licenses that, and the licence looks like a defect. Bootstrapping is Hume's circle read as a feature rather than a bug. Hume said that experience cannot vindicate experience because the vindication would use the method under examination. Reliabilism replies that this vindication is fine, since justification does not require independent access to the reliability fact. That is a coherent position, but it is a position on the far side of Hume's argument, not a development of it. One cannot be both the objection and the answer. Where Hume's own norms are The proto-reliabilist reading is often motivated by a felt need: if Hume has no theory of justification, he seems to have no epistemic norms at all, and that is implausible for a writer who spends a great deal of the Treatise