Philosophy Classics: Philosophy of Science

Popper's target was the Vienna Circle, who used verifiability to distinguish science from metaphysics. Popper reversed it: induction is a myth — no number of white swans proves "all swans are white," yet a single black swan refutes it. So the mark of science is not confirmation but falsifiability: a theory that in principle no observation could ever refute (he named astrology and certain versions of psychoanalysis) is not invincible but empty of scientific content. Science advances by bold conjecture + severe refutation, forever trying to knock down its own hypotheses.

Popper's method is engineered into machine learning: holdout sets and cross-validation are precisely the operationalization of "attempted refutation" — a model that cannot be falsified on unseen data (overfit to the training set) has no predictive value. Null-hypothesis testing shares the root: we never "prove" a hypothesis, only fail to reject it. Falsifiable = testable — this is the watershed between science and rhetoric that "can never be wrong."

Kuhn directly challenged Popper's rationalist picture. Studying the history of science, he found that scientists mostly do "normal science" — puzzle-solving within an established paradigm, setting anomalies aside or patching them rather than discarding the theory. Only when anomalies accumulate into crisis does a "scientific revolution" erupt, one new paradigm wholly replacing the old (geocentric → heliocentric; Newton → relativity). Sharper still is incommensurability: rival paradigms share no neutral observation language to adjudicate between them — the shift is more like a conversion of belief than a compulsion of logic.

Kuhn explicitly borrowed from Gestalt psychology: the same figure reads now as a duck, now as a rabbit — sensory data unchanged while perception restructures as a whole. A paradigm shift is the scientific community's collective gestalt flip, isomorphic with the perceptual bistability neuroscience studies (the Necker cube). It also resembles a phase transition in complex systems: past a critical threshold the whole reorganizes, and it cannot be done "half a step" at a time.

Four thinkers answer the same question from different dimensions — what makes a claim scientific, and what makes knowledge reliable?
· Popper (demarcation): the mark of science is falsifiability, not confirmation — a theory no observation could refute has no scientific content.
· Kuhn (history): science advances by wholesale paradigm shifts, not linear accumulation; anomalies must build into crisis, and the flip is gestalt-like.
· Mohism (verification): every statement must pass three tests — root (is there a basis?), source (can it be verified?), use (does enacting it truly benefit?).
· Nyāya (source): knowledge must first be asked after its origin — perceived, inferred, analogized, or received as testimony?
The two Western thinkers press "where the boundary of science lies and how science evolves"; the two Eastern schools press "where valid knowledge comes from and how to test it." Each yardstick guards one dimension, and together they form an "operating system for knowledge": for any incoming assertion, first use Nyāya for source classification — is it perception, inference, analogy, or relayed testimony? Testimony must be traced to its origin. Then run the Mohist positive screen — basis, evidence, and consequences, gate by gate. Next apply Popper's negative stress test — demand "what evidence would prove this wrong," and downgrade any claim that cannot state its falsification conditions. Finally use Kuhn for framework self-examination — watch whether you are merely patching an old paradigm and dismissing all rival evidence as invalid. In the AI age this pipeline is daily self-defense: what large models mass-produce as "well-founded" speech is mostly testimony with no perception to backstop it (Nyāya), pleasing talk that fails the three tests (Mohism), and all-purpose rhetoric that dares state no falsifier (Popper); and the greatest hidden reef of human-AI collaboration is mistaking fluency within a paradigm for the paradigm's permanence (Kuhn).