Cobra Effect · Science and philosophy
Falsifiability
A claim that nothing could disprove tells you nothing.
6 cards, read aloud in 2:38, with a test and sources.
Vienna, 1919. A young man is surrounded by theories that explain everything.
Karl Popper is 17. Marx explains history. Freud explains dreams. Alfred Adler explains the mind. Whatever happens, each theory has an answer. Popper begins to suspect that this is not a strength.
He takes a case to Adler. Adler explains it without seeing the child.
Popper described a child’s behaviour that did not seem to fit Adler’s theory. Adler explained it at once, in terms of his theory. How could he be so sure? Because of my thousandfold experience, Adler said. And now, Popper thought, your experience is thousand and one fold.
The same year, a theory that could have been killed, wasn’t.
Einstein had predicted that starlight passing the sun would bend by a precise amount. If it bent less, or not at all, his theory was finished. In 1919 Arthur Eddington photographed an eclipse and measured the bend. Einstein had bet the whole theory on one number, and the number came in.
That difference became falsifiability.
A claim is scientific, Popper said, only if something could show it false. Einstein risked everything on a measurement. Adler risked nothing, because no possible child could count against him. A theory that explains every outcome predicts none. Popper published it in 1934, and spent the rest of his life defending it.
All swans are white. How many white swans would prove it?
None. A million white swans cannot rule out a black one. One black swan finishes the claim for good. Evidence for a theory is cheap. Evidence against it is decisive. So the good scientist spends her time looking for the black swan, not counting the white ones.
The test for any claim, in or out of a laboratory.
Ask what this would look like if it were wrong. If the answer is nothing, that is your answer. A forecast that is right whether the market rises or falls has forecast nothing. A manager whose plan is vindicated by success and by failure has no plan. A claim that cannot lose has not said anything.
Sources
- Falsifiability, Wikipedia. Popper’s criterion, the Einstein and Adler contrast, the swans, and the objections philosophers have raised since.
- Conjectures and Refutations, Karl Popper, 1963. The first chapter, Science: Conjectures and Refutations, is Popper telling the Vienna story himself, Adler and the eclipse included. Read that chapter if nothing else.
- Karl Popper, Stanford Encyclopedia of Philosophy. The serious survey of his thought. Falsifiability, the objections to it, and how Popper answered them.
Nearby ideas
- The problem of induction. A long run of past successes cannot prove the next one.
- Confirmation bias. We look for evidence that agrees, and rarely test what could say no.
- The razors. Simple rules for cutting away explanations you do not need.
- The ship of Theseus. Replace every part, and ask whether it is still the same thing.
- The trolley problem. Why harm that follows from a plan feels different from harm that is the plan.
- The placebo effect. Much of what a dummy pill seems to do would have happened anyway.
- The replication crisis. Why many famous findings vanished when others tried them again.
- Cargo cult science. Science can copy every outward form and still miss the honesty.