Weigh disconfirming evidence more heavily than confirming evidence

One strong counterexample outweighs many confirming examples for a universal claim.

Why it works

Popper’s insight is that the logical relation between evidence and hypothesis is asymmetric: modus tollens (if A then B; not-B; therefore not-A) is valid, but induction (many B’s observed; therefore always B) is not. In practice, this means a single, well-documented exception to a rule deserves more evidential weight than ten consistent examples — because the exception could definitively falsify the rule while the examples cannot definitively confirm it.

How to do it

1. When evaluating evidence for a rule or generalization, actively look for documented exceptions.
2. When you find one, ask: "Does this exception require abandoning the rule or narrowing it?"
3. Weight your update to the exception more than proportionally — ask what universe of cases the exception reveals.

Evidence

The logical asymmetry between falsification and confirmation (the problem of induction) is well established in philosophy of science since Hume. Bayesian reasoning formalizes this partially: a surprising disconfirmation typically yields a larger update than an expected confirmation. (mechanistic)

In practice, single exceptions are often measurement error rather than genuine counterexamples; careful judgment about exception quality is required before drawing large updates.

Sources

• Popper (1959), The Logic of Scientific Discovery — asymmetry of falsification and confirmation

Common mistake

Dismissing every exception as "an outlier" or "a special case" — this is an unfalsifiable move that immunizes the belief against any possible correction.