Falsifiability

Falsifiable contrasts with verifiable. A claim is empirically verifiable if possible observation statements logically imply the truth of the claim. If actual observation statements do imply the claim, then it is verified. "This raven is black" verifies "There are black ravens." During the 1930s the logical empiricists of the Vienna Circle proposed verifiability both as a criterion of demarcation of science from nonscience and a criterion of meaning. Their idea was that a statement is meaningful if and only if verifiable in principle, and its meaning is given by its method of verification. For the logical empiricists, only empirically verifiable claims make genuine assertions about the world and are, in this broad sense, scientific. All other claims (metaphysical, religious, ethical, etc.) are cognitively meaningless. In his Logik der Forschung (1934; Logic of Scientific Discovery), Popper replied by rejecting the logical empiricists' concern with language and meaning and by noting that verifiability as a criterion of demarcation excludes scientific law claims and thus the core of science itself. For since a law claim is universal in scope (in simplest form, "All As everywhere and everywhen are Bs"), it cannot possibly be verified: there are always actual or potential instances beyond those so far observed. Yet a universal claim can be falsified by a single negative instance. The first observed black swan refuted the claim "All swans are white." (Law claims of statistical-probabilistic forms are more problematic.) Based on this logical asymmetry of verification and falsification, Popper proposed falsifiability as a criterion of demarcation of science from nonscience, although not as a criterion of meaning. According to Popper, nonscience includes pseudoscience (e.g., Freudian psychology and Marxism) and metaphysics, the one fraudulent, the other sometimes providing a valuable heuristic for science. Many deep scientific problems have their roots in metaphysics, but to be scientific, a claim must take an empirical risk. Moreover, falsifiability, as the ongoing risk of falsification in our world, is a permanent status for Popper. No amount of successful testing can establish a hypothesis as absolutely true or even probable: it forever remains conjectural. That all scientific theories remain falsifiable entails fallibilism, the view that our best epistemic efforts remain open to future revision. There can be no certain foundations to knowledge.

Popper's falsifiability doctrine lies at the heart of his empiricist epistemology and scientific methodology of "conjectures and refutations." The latter, he claims, shows how it is possible to learn from experience without induction from the facts. Previously, empiricism had been equated with inductivism. Popper attacked as question-begging the view that we must arrive at our ideas by induction, that is, by first gathering masses of facts and then gradually detecting regular patterns in them—letting nature speak for herself. Rather, said Popper, we first propose a conjecture to solve a problem, then test the conjecture by trying to falsify it. It is the conjecture that tells us which observations are even relevant. Contrary to the inductivists, it does not matter where our ideas come from, only how we test them. There is no logic of discovery, only a logic of testing.

Since law claims can be falsified but not verified, Popper concluded that the way to truth is indirect, by elimination of falsehood. Hence, error, in the sense of faulty hypotheses, is not a bad thing. On the contrary, it is necessary to scientific progress. "We learn from our mistakes." This is Popper's more extreme form of the nineteenth-century idea that science is a fallible but self-correcting enterprise. Since bold hypotheses that yield novel predictions are risky and hence easier to test, Popper urged boldness. He explicitly forbade, as a form of intellectual dishonesty, ad hoc tinkering to save a hypothesis from falsification. Popper spoke of degrees of falsifiability and attempted, with limited success, to measure both simplicity and the empirical content of a claim (how much it says about the world) in terms of its degree of falsifiability.