David Hume on Induction and Causation

A young Scotsman, writing in the 1730s, noticed something quietly devastating about how we think we know the world — and the cracks he opened have never quite been sealed.

David Hume was twenty-eight when he published A Treatise of Human Nature in 1739. It “fell dead-born from the press,” in his own dry phrase. A decade later he reworked the argument as An Enquiry Concerning Human Understanding, and the press finally noticed. What Hume had done, in plain prose, was take two of the things we rely on most — the assumption that the future will resemble the past, and the idea that one event causes another — and show that neither could be justified by reason alone.

He did not mean this as a stunt. He was trying to do for the human mind what Newton had done for the heavens: find the simple laws beneath the surface. The result was a kind of empirical map of thought — and a problem that philosophy is still chewing on three centuries later.

The problem of induction

Start with a small question. The sun has risen every morning of your life, and every morning of every life recorded. Will it rise tomorrow?

Of course it will, you want to say. But Hume asks: how do you know? Try to write the argument out. It looks something like: “The sun has always risen, therefore the sun will rise tomorrow.” That step from has always to will is doing all the work. What licenses it?

Hume splits human knowledge into two kinds. Relations of ideas are things like geometry and arithmetic — true by definition, knowable from the armchair, but they say nothing about what actually exists. Matters of fact are everything else: the cat is on the mat, water boils at a hundred degrees, the sun will rise. These rest on experience.

Now: any inference from past matters of fact to future ones must assume that nature is uniform — that the future will resemble the past. But this principle is itself a matter of fact. So how do we know it? Only from past experience. And there it is: a circle. We justify induction by appeal to a principle we can only justify by appealing to induction.

This is the problem of induction. It is not a quibble about edge cases. It threatens the rational basis of every empirical generalisation we have ever made.

Causation: the missing connection

The same scalpel cuts a second illusion. Watch one billiard ball strike another. The first stops. The second moves. We say the first caused the second to move. But Hume asks: what, exactly, did you observe?

You saw three things, says Hume. Constant conjunction: events of type A have always been followed by events of type B. Contiguity: A and B were close in space and time. Priority: A came first. That is the entire content of the senses. Nowhere in your visual field was there a fourth thing — a force, a power, a glue — that yoked A to B. The supposed necessary connection is invisible.

This is unsettling. Our entire scientific picture of the world is built on cause and effect. Hume has just pointed out that the central ingredient of that picture is not, strictly speaking, anywhere in our experience.

Custom, not reason

Hume's diagnosis is psychological, not metaphysical. The necessary connection is not in the world. It is in us.

When the mind has seen A followed by B many times, it forms a habit. The next time A appears, the mind slides naturally to expect B. That feeling of expectation — the inner conviction that B must follow — is what we then project back outward and call necessity. Causation, on Hume's account, is at bottom an internal impression of the mind, mistaken for an external fact about objects.

“Custom, then, is the great guide of human life. It is that principle alone, which renders our experience useful to us, and makes us expect, for the future, a similar train of events with those which have appeared in the past.”

This is not a counsel of despair. Hume is no village skeptic, smugly announcing that nothing is real. He calls himself a mitigated sceptic. Of course we go on expecting tomorrow's sunrise; of course we go on treating fire as the cause of warmth. We could not live otherwise. The point is that we do this not because we have proved it, but because we are creatures of habit, embedded in the world we are trying to understand. Reason “is, and ought only to be, the slave of the passions” — and of custom too.

The shadow Hume cast

Few philosophers have provoked as much serious response. Immanuel Kant credited Hume with awakening him from his “dogmatic slumber”; the entire Critique of Pure Reason is, in part, an attempt to answer him. Kant agreed that causation is not read off the world by the senses, but argued that it is something the mind imposes on experience as a condition of having any coherent experience at all. Necessity is real, but it is structural — baked into the way we perceive, not the world we perceive.

In the twentieth century, Karl Popper tried a different escape. Science, he argued, does not really proceed by induction. Theories are bold conjectures, and what we do is try to falsify them. A theory that survives serious attempts to refute it is not proved, only corroborated. This sidesteps the logical problem — you never need to derive a universal from particulars — though many philosophers think Hume's worry simply re-emerges in a new costume: why should past corroboration tell us anything about future performance?

Bayesian epistemology takes the most pragmatic route. Treat induction not as deduction but as probabilistic updating: start with prior beliefs, multiply by likelihoods, get posteriors. The framework is powerful, and underlies most of modern statistics and machine learning. But Hume still lurks. The choice of priors, the assumption that observed regularities will generalise, the trust that the data-generating process is stable — these are not derived. They are, in the honest accounting, custom.

Hume did not break science. He showed where science quietly stands. Every model that fits the past and predicts the future, every neural network trained on yesterday's data and deployed tomorrow, every claim about a law of nature, sits on the same unprovable assumption: that the world will keep behaving in the way it has. We have no proof of this. We have only, in his lovely phrase, custom — the great guide of human life. Three centuries on, that is still about as honest as anyone has managed to be about how knowledge actually works.