Hume suggests that there are two main subjects of knowledge: relations of ideas (roughly, mathematics and logic), and matters of fact. For each of these two subject matters, there are two sources of knowledge: direct and indirect.
|Subject Matter of Knowledge||Direct Knowledge||Indirect Knowledge|
|Relations of Ideas||
(some logical or mathematical truths are so basic and obvious that there is no need to derive them from something more basic. Mathematically, these are the kinds of things one wants as axioms. (This is pretty much the same thing Descartes meant by "clear and distinct ideas" and Spinoza and Locke meant by "intuitive knowledge")
(this is what we might call "deduction" or "proof". The idea is that you can prove mathematical truths on the basis of the basic axioms by a series of deductively valid steps.)
|Matters of Fact||
(some things you can just see (hear, etc.) are true, without needing to employ reasoning to make inferences. Like Locke, Hume also includes things we remember having observed in this category.)
(the reasoning that enables us to reach conclusions about matters of fact that we don't directly observe involves what Hume calls "experimental inference." His main question in "Skeptical Doubts Concerning the Operations of the Understanding" is what this is and how it can be justified.)
The issue Hume is mainly concerned with in Enquiry Concerning Human Understanding, section IV, is the nature and justification of "experimental inference."
1. "experimental inference," i.e. drawing conclusions about things we haven't observed on the basis of things we have observed, depends on knowledge of cause and effect. How do I know someone else is on the island?
1: there's a human-footprint-shaped-thing in the sand (I know this
directly, by observation)
2: human-footprint-shaped-things are caused by the feet of humans
(this is a general causal principle)
3: there is a human on the island with me
2. But now the question arises, how do I know #2 above? #2 is a claim, not
only about things I have observed, but also about cases I have not observed (otherwise it would be of no use in trying to draw conclusions about things
I have not observed, like the presence of someone else on the island).
Knowing that footprints have been caused by people in the cases I have already observed does not help me any in drawing conclusions about things I haven't observed unless I have some reason to think the cases I haven't observed will be like the ones I have observed.
So how can I justify my claim to know this general causal principle? Well, given Hume's classification of sources of knowledge, there seem to be four possibilities (the same ones I just described a couple of paragraphs ago:
(a) intuition. But I intuitively know something only if I see that it has to be true, that it could not fail to be true. However, causal laws are not necessary in this sense; there is no problem in imagining that they might fail to be true.
(b) demonstration. But this has a similar problem to that faced by intuition. Proving something essentially involves showing that it couldn't fail to be true (as Hume puts it, showing that it's opposite involves a contradiction). But there is no contradiction in supposing that a causal generalization won't apply in a future case.
(c) observation. But we're considering how to justify knowledge of something I haven't observed, so obviously observation will be of no help.
(d) there's only one possibility left, assuming that Hume's classification is correct, and that is experimental inference. But our initial question was what justifies experimental inference! To say that it is justified by experimental inference is to offer a completely circular justification.
NOW, how does all this relate to induction and how we know that the future will be like the past? "Induction" pretty much means the same thing as "experimental inference," so Hume's question about how to justify experimental inference could also be described as a question about the justification of induction.
And the question of whether the future will resemble the past is a particularly jazzy example of the attempt to apply our knowledge of things we have observed to things we have not observed: in this case, to determine, on the basis of past observations, what the future will be like. (It's not the only such example; one could make the same points by considering, for example, how a few hundred observed answers to an opinion poll can justify inferences about the beliefs of millions of unobserved people.)