What, if anything, is established by Hume’s sceptical argument about induction?

Submitted for PH201 (History of Modern Philosophy) Summer Exam

In this essay, I consider Hume’s argument about induction, and argue that although he succeeds in demonstrating that the process of inductive reasoning he describes is not epistemically justifiable, he does not succeed in establishing grounds for scepticism because such reasoning is not required to obtain beliefs about unobserved matters of fact.

Hume believes that the objects of human inquiry can be distinguished into “Relations of Ideas”, such as mathematical objects, knowledge of which is a priori and certain; and “Matters of Fact”, which are known a posteriori on the basis of experience, since they depend on what is actually existent in the world. Since knowledge of matters of fact is obtained via experience, he addresses the question of how we arrive at beliefs about that which we do not observe, such as our belief that the sun will rise tomorrow. He claims that the basis for such beliefs is inference from that which we have observed, through induction: we believe that the sun will rise tomorrow because it has risen on every previous day. Of course, such an inference must be justified, and he claims that the justification for such inferences is founded on the principle of cause and effect: like causes lead to like consequences; when a billiard ball strikes another, it causes the movement of the other, so we can infer that when the same thing happens in future, the same causal relationship will again cause similar movement. However, Hume argues that the relation of cause and effect is not one we ever experience directly; all we ever experience is the constant conjunction between various kinds of events. There is, he claims, no legitimate argument from the fact that in our experience, some event has typically been accompanied by another, to the view that it will always be so accompanied in future. The best one can do is appeal to the uniformity of nature, that “the future will be conformable to the past”; but this cannot be justified inductively, for the argument would be circular; and nor can it be demonstrated through reason. Thus, inductive reasoning has no rational foundation; it is a mere custom, or habit.

Against this, it can be argued that Hume has too high a standard in mind. In particular, all he has shown is that there is no deductively valid argument for the inference from observed matters of fact to unobserved matters of fact. But as Strawson argues, “deduction is one kind of reasoning and induction is simply a distinct, fundamentally different kind of reasoning.” There is no sense in thinking that the latter must conform to the standards of the latter; the question is whether inductive inferences of the kind we frequently make are inductively justifiable, which they trivially are. Moreover, there is a good pragmatic reason for believing that we are on sound footing in making inductive inferences: Hume himself concedes that “Custom … is the great guide of human life”, and since we cannot do without inductive reasoning, it remains the best we can do in discovering what’s true.

However, neither of these arguments satisfactorily deal with the problem of induction. As Laurence Bonjour puts it, “it seems to be suggested (though often not very explicitly), the skeptical challenge is adequately dealt with, even if not exactly refuted” (emphasis mine). To argue as Strawson does that inductive inference is almost definitionally justified because it is the cornerstone of inductive reasoning does not tell us anything about what is or is not the case. Matters of fact, in Hume’s terminology, are true or false on the basis of what actually exists in the world; rephrasing our epistemic standards does not change that. If indeed it is true, as the pragmatists argue, that we cannot help but rely on induction, then either there is some identifiable reason that it is a useful guide to navigating reality (in which case, we must identify the reason), or else we are back to Hume’s view that although we cannot help but rely on it, there is no rational case for it; after all, beliefs arrived at through induction sometimes turn out to be false.

It has been suggested, for instance by Beauchamp and Rosenburg, that Hume did not mean to advocate the sceptical conclusion, and only demonstrate that induction did not attain the epistemic rigour of deductive reasoning. One reason for thinking this is that he himself uses inductive reasoning in his philosophy of natural religion, arguing that the reason one typically believes people’s testimony is because one usually finds that it has been correct in the past. But for the reasons given, this is not tenable: the argument against induction does lead one to scepticism about conclusions one thereby obtains, unless it can be defeated.

A part of Hume’s argument with which one might take issue is his account of causation. His inductive scepticism depends on the fact that if we consider our experience, we never actually “find” a cause. If such causal relations could truly be observed, then we would be able to know the effects of some events a priori, since to say “A causes B” is to say that “A necessarily leads to B”. But whether we are observing external bodies or our own internal decisions, it is only ever through experience that we learn the effects of an event. Thus we can never get this impression of necessity which must be the concomitant of causation; all we ever find is mere “succession” of events.

Anscombe has criticised Hume’s introspective method as motivated reasoning: “It is argued that ‘all we find’ is such-and-such, and it turns out the arguer has excluded from his idea of ‘finding’ the sort of thing he says we don’t ‘find’”. Hume is perfectly content to say we perceive, say, a billiard ball as a individuated substance (and not, say, a collection of qualities); yet he denies that we find causality. Thus she argues that this does not demonstrate what Hume thinks it does. Furthermore, Hume’s account of causality is that it is events which cause other events; but this is not self-evidently the idea of causality we possess. For instance, in Kant’s conception, causes are substances (or their causal powers), and effects are the changes of state in those or other substances. If we adopt such a model, it arguably transpires that we do in fact observe causality in experience, and this may provide a foundation for induction.

Taken together, these criticisms show that Hume’s sceptical argument fails, because it does not rigorously dispense with causal realism. Alexander Miller uses the example of a billiard ball which collides with another; then the second billiard ball moves away, and simultaneously, an onlooker sneezes. The point is that the collision caused the movement, but not the sneeze. On Hume’s account, this is because there is a constant conjunction in the one case and not the other; usually people don’t sneeze here. But supposing in someone’s experience they always did, perhaps because they have an allergy to chalk. Then it would be possible to form an explanation of the conjunction, as a state-transition precipitated by a causal power of that element. This hypothesis implies a prediction for what will happen in the future; the conditions of this “experiment” can be varied. Thus, we arrive at a Popperian view: as David Deutsch puts it, “inductivism rests upon a mistaken idea of science as seeking predictions on the basis of observations, rather than as seeking explanations in response to problems”. We obtain knowledge about unobserved matters of fact by having an explanatory model of the causal relationships at play in the world, and the explanation enables us to make predictions.

Thus, Hume’s sceptical argument dissolves. It is true that inferring that the future will be like the past is invalid, and Hume demonstrates the problem with such reasoning. But if one reconceives of the nature of causality as discussed, and of the nature of reasoning used in forming beliefs about unobserved matters of fact, the grounds for scepticism disappear.

Result

Mark: 74% (Low First), averaged across this and two other answers.