Why does a scientific theory have to be able to fail?

A scientific theory has to be able to fail because a theory that no observation could ever contradict isn't telling you anything about the world. That's the idea of falsifiability, introduced by the philosopher Karl Popper in 1934: a claim is falsifiable if you can describe an observation that would prove it wrong. Take "all swans are white". You could never verify it, because you'd have to see every swan. But one black swan is enough to refute it.

Popper got the idea from real life. As a young man in Vienna, after training as a cabinetmaker, he volunteered in one of the psychoanalyst Alfred Adler's clinics for children. He noticed something odd: whatever a person did, Adler's theory could explain it. That looked like strength, but Popper decided it was a weakness. A theory that explains everything excludes nothing, so it predicts nothing.

Einstein's relativity was the opposite. It stuck its neck out, predicting that gravity bends light, something Newton's physics didn't, and in 1919 Eddington's eclipse measurements could have sunk it. They didn't. For Popper, surviving a risky test like that is what makes a theory worth trusting.

Falsifiability isn't a perfect sorting machine. Any test also relies on side assumptions, so a failed prediction doesn't always tell you which part is wrong. And "unfalsifiable" doesn't mean "meaningless", only "not science". But the core question still works anywhere: what would change your mind?