Every evening, millions of us set an alarm in the unshakeable confidence that the sun will rise. We drink water expecting it to quench our thirst rather than poison us, and step onto bridges assuming they will bear our weight, as they have borne the weight of countless others. None of these expectations is unreasonable. Yet in the eighteenth century the Scottish philosopher David Hume asked a question so simple that it is easy to miss how unsettling it is: what, exactly, justifies them?
Our expectations rest on induction: reasoning from observed cases to unobserved ones, from 'every swan I have seen is white' to 'all swans are white', or from 'the sun has always risen' to 'the sun will rise tomorrow'. Unlike deduction, in which true premises guarantee a true conclusion, induction always leaves a gap. No number of white swans logically entails that the next one will be white – as Europeans discovered when they reached Australia and came across black swans.
Hume's insight was that the gap cannot be closed in the obvious way. We might say that induction is justified because nature is uniform: the future resembles the past. But how do we know that nature is uniform? Only because, in the past, it always has been. We would be using induction to justify induction, an argument that moves in a circle. Nor can the principle be proved by pure reason, since there is no contradiction in supposing that the course of nature might change. We can coherently imagine, as Hume himself suggested, bread that ceases to nourish us, or a stone that falls upwards; we simply do not expect them.
Bertrand Russell gave the problem its most memorable illustration. A chicken, he observed, which has been fed every day of its life by the farmer, comes to expect food whenever the farmer appears. Its expectation is perfectly well founded, right up until the day the farmer arrives for a different purpose altogether. The point is not that we are as foolish as chickens, but that the chicken's reasoning has exactly the same form as ours. Experience, however long, cannot by itself tell us when a pattern is about to break.
Hume's own response was strikingly calm. He did not suggest that we should stop expecting the sun to rise. Reason, he argued, cannot justify induction, but custom or habit compels us to rely on it, and we could not live otherwise. Nature, as he put it, is too strong for principle. The sceptic may win the argument in the study, but he will still duck when a ball comes flying towards his head.
In the twentieth century Karl Popper proposed a bolder escape. Science, he claimed, does not rely on induction at all. Scientists put forward conjectures and then try as hard as they can to refute them. A theory that survives repeated attempts at falsification is not thereby proved, but it has earned the right to be taken seriously for the time being. Popper's account captured something important about the spirit of science: its readiness to be shown wrong. Critics, however, noted that when we choose to build a bridge on a well-tested theory rather than an untested one, we seem to be assuming that past success is some guide to future reliability – which is induction by another name.
Nelson Goodman later showed that the difficulty runs deeper still. He invented the word 'grue', meaning green if examined before a certain future date and blue if examined after it. Every emerald examined so far is green, but it is equally grue, so the same evidence 'supports' two predictions that will diverge sharply on that date. Why is it reasonable to project 'green' into the future but not 'grue'? The answer cannot simply be that green is the simpler word, because simplicity seems to depend on the vocabulary we happen to start with.
None of this means that scientists are fools or that we should distrust our bridges. 'The problem of induction is not a reason to stop reasoning,' says Dr Owen Llewellyn, a philosopher of science at a university in Aberystwyth. 'It's a reminder that our knowledge of the world is provisional. The best scientists already behave as though that were true: they hold their theories firmly enough to act on them and loosely enough to give them up.' There is a fitting irony here. Our confidence in that lesson rests on the very kind of reasoning it calls into question: the history of science has been confirming it, case by case, for centuries.