Falsifiability

Falsifiability is the criterion proposed by Karl Popper to demarcate scientific hypotheses from non-scientific ones. A hypothesis is falsifiable if there exists, in principle, an observation or experiment that could prove it false. Theories that cannot be falsified — that accommodate any possible outcome — are not wrong. They are not even scientific.

The criterion is a direct consequence of taking seriously the asymmetry between inductive and deductive logic: no finite number of confirming observations can prove a universal hypothesis, but a single well-established counter-instance can refute it. Science progresses not by accumulating verifications but by surviving attempts at refutation. A theory that has repeatedly been exposed to falsification and survived is corroborated — but corroboration is not proof. It is the absence of disproof.

Falsifiability has been controversial since its introduction. Critics note that well-entrenched theories are rarely abandoned on the basis of a single anomaly — the history of science is full of anomalies that were eventually explained within the existing framework. Thomas Kuhn argued that scientific revolutions follow a social and historical pattern that Popper's criterion does not capture. Lakatos developed the concept of research programmes to accommodate the reality that scientists rationally protect theoretical cores from falsification.

The deeper question falsifiability raises is not demarcational but epistemological: what kind of evidence should change our minds? This is the question Bayesian Epistemology attempts to answer with more precision.