Hume’s argument for inductive skepticism is contained in section 4 of the Enquiry. The topic of this section, as Hume states it, is “the nature of that evidence, which assures us of any real existence and matter of fact, beyond the present testimony of our senses, or the records of our memory” (4.1.3). Thus, like Millican, I will mean by “induction” all factual inference to the unobserved rather than merely predicting the future. Hume clearly intends his argument to cover all inferences about unobserved objects, even when those objects presently exist.
The first premise in Hume’s argument is that induction is based on knowledge of cause and effect. This includes inferences about contemporaneous but unobserved objects, since those depend on the object’s past causal history and the assumption that what is presently happening to the object is predictable based on cause and effect. As Hume first states the premise, it is ambiguous: He says, “All reasonings concerning matter of fact seem to be founded on the relation of Cause and Effect” (4.1.4). He might mean that induction is based on the relation of cause and effect itself rather than knowledge of it, and then there would be no skeptical conclusion, since while he proceeds to show that knowledge of the relation is not based on reason, he continues to believe in the relation itself. However, the next paragraph makes it clear that he means knowledge of the relation. He says there, “If we would satisfy ourselves, therefore, concerning the nature of that evidence, which assures us of matters of fact, we must enquire how we arrive at the knowledge of cause and effect” (4.1.5) The skeptical route is still open.
The next premise is that knowledge of cause and effect has four possible rational sources: intuitive evidence, sensory evidence, demonstrative argument, and probable argument (Millican). Hume rules out the first two sources on the grounds that “[n]o object ever discovers, by the qualities which appear to the senses, either the causes which produced it, or the effects which will arise from it; nor can our reason, unassisted by experience, ever draw any inference concerning real existence and matter of fact” (4.1.6). This seems obviously right: we cannot discern the causal properties of an unfamiliar object simply by considering the object itself. If it seems that we can find out the causal properties of familiar objects just by thinking or sensing them, this is only because we can draw on our past experience with the object or objects like it. Hume next rules out demonstrative argument on the grounds that in an argument with a conclusion about cause and effect, there is no logical contradiction involved in supposing the conclusion to be false (4.1.5).
Hume finally attempts to rule out probable argument as a possible source. He argues, “all our experimental conclusions proceed upon the supposition that the future will be conformable to the past. To endeavour, therefore, the proof of this last supposition by probable arguments, or arguments regarding existence, must be evidently going in a circle, and taking that for granted, which is the very point in question” (4.2.6). Below, I will take issue with this part of Hume’s argument. At the moment, I will simply point out that regarded in this way, the argument clearly has a skeptical conclusion. Hume recognizes four possible rational sources for knowledge of cause and effect and tries to rule out all of them. He regards these four possibilities as exhaustive: “it may be requisite to venture upon a more difficult task; and enumerating all the branches of human knowledge, endeavour to show that none of them can afford such an argument” (4.2.4). His conclusion is that we have no knowledge of cause and effect, and thus that the inference between “the present fact and that which is inferred from it” is “entirely precarious (4.1.4). These are clear normative terms.
Even independent of this textual evidence, the cognitive science reading of the argument fails because it either collapses into the skeptical reading or makes Hume’s argument look embarrassingly bad. On the cognitive science reading, Hume is concerned to show what causes our factual inferences to the unobserved, not what could justify them. Hume shows that any argument for the conclusion that causation is real, or that some X caused some Y, would be unsound. According to the cognitive science reading, he then concludes from this that no argument could cause us to form a belief in causation. However, if this were the case, then Hume would rely on the assumption that we never form beliefs based on unsound arguments and yet leave this assumption totally unargued for. The assumption also seems false, since people often accept conclusions on the bases of arguments they do not realize are unsound, and conversely, fail to accept conclusions of sound arguments. Even if Hume did justify this assumption, so that his argument were not so bad, the reading would then collapse into the skeptical one since Hume would still be making use of the normative concept of a “good argument.” Thus, whether he liked it or not, Hume’s argument would have a skeptical conclusion, since it would show that factual inferences to the unobserved do not rest on any good arguments. The cognitive science reading is both out of line with Hume’s intentions and uncharitable, since it makes Hume’s argument vulnerable to an obvious and true objection.
Hume, then, intends his argument to be skeptical. However, we ought not to find the argument troubling, because it does not quite succeed in establishing its skeptical conclusion. The error lies in Hume’s treatment of probable arguments. Recall that probable arguments form one of the four possible sources for rational justification of factual inferences to the unobserved. Hume rejects the possibility of justification through probable argument on the grounds that any probable argument in this case will rely on the uniformity thesis. However, it is not clear that this is true. For instance, take the following probable argument:
First premise: This is bread.
Second premise: Bread has nourished me in the past.
Conclusion: Therefore, this bread will nourish me.
Clearly this argument is not deductively valid, since the premises do not deductively entail the conclusion. Hume recognizes this and acknowledges that simply ruling out deductive arguments is insufficient; probable arguments are inductive, not deductive. Anyway, even granted the further premise that the future resembles the past (the “uniformity thesis”), the argument would be deductively invalid, since this premise as Hume states it would be too vague to guarantee the conclusion (Millican 152). Hume recognizes that he must do something more: he must show that the argument is invalid according to the standards of induction, too.
However, Hume fails to meet this further burden. The above argument is a perfectly valid inductive argument. It does not require the uniformity thesis as a further premise. At the least, Hume would have to say more to explain why the argument does require the uniformity thesis in order to be inductively and not just deductively valid. Yet I do not see how he could. Induction is simply the practice of making inferences based on past experience, just as deduction is the practice of making inferences based on certain rules like modus ponens. Within the practice of deduction, we presuppose rules like modus ponens; they are not themselves in need of deductive justification. If they were, no deductive argument could ever get going; we would be stuck with an infinite regress like that in Carroll’s “What the Tortoise Said to Achilles”. The case is similar with induction; we presuppose in induction that the past is a reliable guide to the future. This thesis is not a further premise that requires justification.
Given that he cannot appeal to the standards of inductive justification (since they affirm the validity of inductive arguments) nor to the standards of other practices like deductive justification (since they are irrelevant for his purposes), it seems that Hume has only one possible defense. He must reply that in ruling inferences like the above to be unjustified, he is appealing to the standard of all justification whatsoever. In that case, I have two responses. First, Hume has not convinced me that the argument fails to meet this standard, whatever it may be, since the bread argument and those similar to it seem to me to be perfectly justified. Second, and more radically, I object that I do not understand what Hume means by “the standard of all justification whatsoever.” I would understand what he meant by “the standard of deductive justification” or “the standard of inductive justification” or, more generally, any standard of justification relative to a practice. But someone who appeals to the standard of all justification whatsoever tries to transcend all practices. Here I cannot see any place for questions of justification. Induction is a practice we all engage in, and one, as Hume recognizes with his talk of custom and habit, that is a natural part of human life. It is justified relative to the standards of the practice of induction, and what more is there to say? Hume and I might at this point be at an unbridgeable impasse, but like anyone using terms his audience does not immediately understand, Hume has the positive burden here to show what he means by “justification.” Even if justification is not, as a matter of principle, always internal to a practice, Hume has at least not met his burden in this case. I am willing to reflect on the practice of induction, too. I can step back with Hume and recognize that the practice of induction is not founded on any further reasoning but rather on custom and habit. Yet Hume has not done enough to show why this ought to make us think that induction is not justified.
Further, there is another reason, independent of the above objection, that Hume’s argument should trouble us less than it might seem. Even if Hume succeeds in establishing his skeptical conclusion, his argument does not prove induction is unjustified. It proves merely that we have no reason for believing it to be justified. This is because Hume’s own argument depends on a factual inference to the unobserved. He reflects on past instances of induction and concludes that since he can find no rational argument supporting any of them, no inductive arguments can be founded on rational argument. As Hume has himself shown, this conclusion is not founded on any rational argument. There is no reason to believe that future inductive arguments will resemble past inductive arguments. Thus, there is no reason to believe that they will not be justified; perhaps they will be justified by some argument we have not yet discovered. On its own, this defense of induction is fairly weak. Hume would still have shown that all past factual inferences to the unobserved have been unjustified, a quite troubling conclusion on its own. Most of the weight of my defense of induction must rest on my first objection.
Hume and I agree on everything but final conclusions. He has shown something profound about the place of reason in our lives (with regard to induction specifically, though I think the point could be extended to any kind of reasoning). I just do not see why the final conclusion ought to be skeptical. This is why the argument does not trouble me.