Falsifiability

Foundation · 12 min

A claim that explains everything predicts nothing

The criterion made famous by the philosopher of science Karl Popper is this: a claim carries testable information if it can say which observation would prove it wrong. Consider the sentence “Be careful today — it's a day open to surprises.” If the day brings mishaps, the reading seems confirmed; if it passes smoothly, it seems confirmed again — “because you were careful”. If no possible day can prove the sentence wrong, then the sentence is actually saying nothing about the day.

That does not make the sentence malicious; but it does mean it should be treated as conversation or comfort, not as a source of information.

A reading that explains every outcome is like a bet that wins on both heads and tails: because it takes no risk, it also earns nothing.

An example from science: the risky prediction

Einstein's general theory of relativity predicted that light from distant stars would bend by a specific amount as it passed near the Sun. The prediction was put to the test during a solar eclipse in 1919. Here is the crucial point: had the observation turned out differently, the theory would have been in serious trouble. That is exactly what made the theory strong — it dared to enter an exam it could fail, and it passed. An explanation that makes no risky predictions and accommodates every outcome enters no exam at all, and therefore can never be said to have passed one.

Falsifiable does not mean false

The word may sound negative, but falsifiability is a virtue, not a defect. A falsifiable claim is one that says “here is how you may test me”; it does not mean the claim is false. The Moon's phase cycle is a falsifiable claim, and every observation ever made has supported it — that is precisely what strong knowledge looks like. This is why scientists use “never yet falsified” as praise for a theory: it has entered the exam again and again, and stayed standing every time. A claim closed to testing, by contrast, can neither be proven wrong nor genuinely supported; it stands outside the scales of evidence altogether.

Escape hatches

If a new exception is added every time a claim fails — “another influence was dominant that day”, “your case was special” — the claim gradually becomes untestable. Each exception may look reasonable on its own; the pattern must be viewed as a whole: can this claim ever turn out wrong? If the answer is “no — every outcome gets explained somehow”, what you hold is no longer a testable claim but an explanation that fits any shape. A practical criterion you can apply to yourself: if you have added more than one exception to an idea you defend within the past year, it may be time to check whether that idea is still testable. This question must also be turned on our own ideas; in the next lesson we will meet the mechanism built for exactly that purpose: blind assessment.