Or, why empiricism isn’t as crazy as Hume thought
Awhile back I promised a post on “scientism”–a popular target of religious apologists, though it’s not always clear what it’s supposed to be. Before I discuss it, though, I want to discuss a more philosophically respectable idea, empiricism, and the very philosophically respectable objection to it: that it cannot serve the problem of induction.
Quick definitions:
(1) Empiricism: The doctrine that outside a special class of conceptual truths, which are true in virtue of how the concepts involved are defined, all human knowledge depends on evidence and observation
(2) Induction: The process of inferring general facts about the world from many specific facts. For example, inferring “the sun rises every day” from “the sun rose today, yesterday, two days ago, three days ago…” or inferring “everything orbiting the Sun follows Kepler’s laws” from “Mercurey, Venus, Earth, and Mars follow Kepler’s laws.
(3) Problem of Induction: In the words of the SEP, the question of “can induction be justified?”
David Hume, possibly the most famous empiricist ever (though he’s got competition from John Locke), thought that because empiricism is true, induction can’t be justified, and since induction can’t be justified, we can’t know things like “the sun rises every day,” and the therefore we can’t know things like “the sun will rise tomorrow.” Other philosophers, such as Laurence Bonjour, have reversed the argument: if empiricism were true we couldn’t know that the sun will rise tomorrow, but plainly we can know this, so empiricism is false.
For a year or two of my undergraduate career, I thought this argument was totally conclusive, and wondered why everyone didn’t accept it. But why think induction couldn’t be justified if empiricism were true? Hume put it this way: “all inferences from experience suppose, as their foundation, that the future will resemble the past.” This is no conceptual truth, since we can conceive of the future bearing no resemblance to the past. But it can’t be known by experience, either: “It is impossible that any arguments from experience can prove this resemblance of the past to the future; since all these arguments are founded on the supposition of that resemblance.” To prove “this resemblance” by experience, then would be circular reasoning.
Richard Chappell has endorsed an argument somewhat like this, but ironically, I credit Richard with persuading me that the argument is a bad one. The trouble is this: not all elements of an inference are on equal footing, something Richard has pointed out with respect to intuitions and pesuppositions. In the case of intuitions: lots of philosophers are convinced that no chain of reasoning can get started without accepting that some things are just obvious or “known by intuition,” but that doesn’t mean that the obviousness of an argument’s premises is itself a premise. Similarly, to mathematical theorem means assuming that, at each step, you can accurately remember what you did in the previous parts of the deduction, but that doesn’t make “my memory is reliable” an important mathematical axiom.
Similarly with induction. Suppose there is a Principle of Justified Induction that says “under such-and-such circumstances, being justified in believing that a generalization holds for many cases justifies believing that the generalization holds for all cases,” and that principle is true. If such a principle were true, then inferring “the sun rises every day” from “the sun rose today, yesterday, two days ago, three days ago…” would, under the such-and-such conditions, be a good inference, and the Principle of Justified Induction would not play the same role in the inference as the belief that “the sun rose today” played. Neither would Hume’s statement that “the future will resemble the past.” The trouble is that normally, it’s beliefs that need justification, while induction is a process, and expecting a process to be justified the way a belief is is a recipe for a headache.
If this isn’t obvious, I suggest reading Lewis Carroll’s short story “What the Tortoise Said to Achilles,” which makes the same point about deduction. Were Hume’s criticism of deduction legitimate, we could also conjure up a “problem of deduction” by claiming that all deductions rely on the supposition that deductive inferences work, and then ask where that supposition comes from.
Interesting corollary: if I’m right about all this, the seemingly odd suggestion that we can inductively justify induction is half-right. It’s not that we really can do this, but that we can inductively justify beliefs about induction. Since beliefs about induction are not part of inductive inferences the way instances of the generalization to be justified are, there’s no risk of circularity at least in the ordinary sense of the word.
Obligatory paradoxical epilogue: I’m sure I’ve missed something here. Any philosophy nerds reading this (Richard? You reading this?) please tell me what it is.
While it may be true that ‘induction cannot be justified by empirical methods’ it would be foolish to bet against it. If history proves anything it is that induction is more likely to lead to true beliefs about the world than any other method.
“The trouble is that normally, it’s beliefs that need justification, while induction is a process, and expecting a process to be justified the way a belief is is a recipe for a headache.”
I’m always happy to see more people recognize the distinction between premises and inferential methods. But we need to be careful about what conclusions we draw from this. For you are implicitly neglecting the following principle, which strikes me as pretty self-evident:
(Methods We Can Believe In) If it’s rational to infer Q from P, then it’s rational to believe this epistemic principle: that belief in P rationally licenses belief in Q.
So, for example, we can be justified in our inductive reasoning only if we can also be justified in believing the Principle of Justified Induction — which in turn is presumably justified only if we’re also justified in believing that the future will resemble the past.
This is no “problem”, contra Hume, because we’re perfectly justified in believing that the future will resemble the past. Still, it’s clear that this must be a priori justifiable if it’s justifiable at all. So it’s a priori justifiable. So empiricism is false.
If we are talking about Hume himself rather than later constructions of him, Hume does think there is a ‘problem of deduction’, or something analogous to it; that’s the point of Treatise 1.4.1. But I think Hume’s not actually concerned with justifying induction, as such (or usually even with justifying anything or denying that anything is justified); he’s interested in our reasons for using the principles we use in our causal reasoning: the same causes have the same effects, the future will resemble the past, etc. And his argument is that these are ultimately grounded in practical necessity: we find that we’re set up in such a way that we think in those terms whether we want to do so or not, and that a lot of times we can get reasoning that uses such principles to work quite well for practical purposes. Arguments and necessary principles don’t really play a role. And that’s not surprising; it’s twentieth and early twenty-first century analytic philosophers, not early modern empiricists, who think justification is an interesting subject. Hume is interested instead in the question of the relation between inquiry and human nature: what in fact is the reason we engage in induction in the way we actually do, and how that relates to other aspects of human life and inquiry.
proudfootz: The problem is that the inference from “induction has more often led to true beliefs about the world than any other method” to “induction will, in the future, lead to more true beliefs about the world than any other method” is itself an inductive inference. We need to know if that inference can ever be justified.
Richard: when writing the post, I suspected some principle like that might be invoked here, though I wasn’t sure how it would be articulated. You’ve given a nice attempt at formulating such a principle. However, consider: does this principle mean that all facts about what beliefs license what other beliefs are rational to believe under all circumstances? If so, the principle seems to me to be false, but if the principle doesn’t require anything so extreme, then why not? What does it entail?
Brandon: 17th and 18th century philosophers didn’t toss around the word “justification” as much as today’s philosophers do, but they did seem to be concerned with it, or at least something an awful lot like it: Descartes talked about trying to get assurance our beliefs are correct and finding a way to resolve disputes, Locke wanted a way to resolve disputes and seemed to think knowing how we know things would help with that, and Hume was alternately bothered by his own conclusions about induction or hoped that they would somehow quiet fanatics.
Chris – Right, some people may be so confused, or beset by other false beliefs, that they can’t rationally believe what’s objectively warranted and knowable by agents in good or teleologically ‘normal’ epistemic conditions. But this indicates the way to explicate the principle: for example, we could simply invoke the weaker claim that true epistemic principles must be generally knowable, or something along those lines.
Chris,
In other words, a lot of things that don’t have to be considered together, or even thought particularly relevant to each other, interested the philosophers of the seventeenth and eighteenth centuries; and we happen to put them together under the phrase ‘justification’. Each of the early modern philosophers was setting out to do something rather different from anything done by modern epistemologists; any similarities are simply developed in a way incidental to these main aims. I think we should be skeptical of attempts to read Descartes, Locke, and Hume as very interested in justification (I think we should be skeptical of the value of thinking in terms of justification, for that matter, for precisely the reason that our use of the term tends to oscillate between a mish-mash of very different things and very narrow technical uses that only cover a tiny portion of epistemic life; but that’s another argument).