Philosophy of science for beginners: what makes a claim scientific?

I work on quantum many-body systems in Copenhagen, and like most physicists I spent years doing science without asking what science is. Then a friend asked me, over dinner, why I trusted my results and not her horoscope. I gave a confident answer, and by dessert she had found three holes in it. This guide is what I wish I had read before that dinner.

Philosophy of science asks what distinguishes science from other enterprises, how scientific claims are justified, and what successful theories tell us about the world. Here are its main questions, in the order a beginner usually meets them.

Where does science begin?

The story many of us were taught has experimental science born in seventeenth-century Europe, with Francis Bacon calling for knowledge built from careful observation and warning against the "idols of the mind", the habits of thought that distort it. That story leaves things out. In the early eleventh century Ibn al-Haytham, working in Cairo, built apparatus to test competing theories of vision, varied conditions deliberately and checked hypotheses against the results; his Book of Optics became a standard work in Latin translation. His contemporary al-Bīrūnī analysed errors in measurement systematically. Claims that Ibn al-Haytham single-handedly "invented the scientific method" overstate the case, but the narrower point stands: systematic experiment is centuries older than the textbook story says.

The demarcation problem

The first big question is how to draw the line between science and non-science. It matters practically: courts, schools and funding bodies draw it every day.

The logical positivists of the early twentieth century proposed verification: a claim is scientific, indeed meaningful, only if experience could confirm it. The trouble is that no universal law can be conclusively verified. However many metals you heat, "all metals expand when heated" goes beyond your evidence.

Popper and falsifiability

Karl Popper turned the criterion around. A theory is scientific, he argued, if it forbids something, if some possible observation would refute it. "All swans are white" is scientific because a black swan would sink it. Popper's targets included theories he thought could explain any outcome whatever. A theory compatible with every possible observation, he said, tells you nothing about the world.

This is a powerful idea, and it explains why a bold prediction that might fail is worth more than a tidy explanation of what has already happened. It was the answer I gave my friend about the horoscope. Then she asked what happens when a prediction fails.

The Duhem–Quine problem

Here is the catch, identified by the French physicist Pierre Duhem and later generalised by W. V. O. Quine. Theories are never tested alone. To get a prediction you need the theory plus a bundle of auxiliary assumptions: the instrument works, the sample is pure, nothing unknown interferes. When the prediction fails, logic tells you something in the bundle is wrong. It does not tell you what.

The history of astronomy gives the perfect pair of cases. In the nineteenth century Uranus did not move as Newtonian mechanics predicted. Astronomers kept the theory and postulated an unseen planet, and Neptune was found. Mercury's orbit also misbehaved. Astronomers tried the same move and postulated a planet, Vulcan, which does not exist. Mercury was eventually explained by general relativity. Identical reasoning, opposite outcomes, and nothing in the logic of falsification told anyone in advance which was which.

Kuhn and Lakatos

Thomas Kuhn's The Structure of Scientific Revolutions (1962) made the historical point dramatic. Most science, he argued, is "normal science" carried out within a paradigm: a shared set of theories, methods and exemplary problems. Anomalies are tolerated or explained away until they pile up into a crisis, and then a revolution replaces the paradigm, as Copernican astronomy replaced Ptolemaic. Kuhn's critics accused him of making science irrational, a matter of conversion rather than evidence. He resisted that reading.

Imre Lakatos offered a middle path. Scientists work within research programmes with a protected hard core and a belt of adjustable auxiliary hypotheses. A programme is progressive when its adjustments predict new facts, as Neptune did; degenerating when it only patches holes after the fact, as Vulcan eventually showed. That gives you a way to judge rescue moves without pretending a single failed prediction settles everything.

Are electrons real?

A second big question concerns what theories tell us. Scientific realists say our best theories are approximately true and the unobservable things they describe, electrons and fields, really exist. Their main argument, associated with Hilary Putnam, is that the success of science would otherwise be a miracle.

Anti-realists reply with history. Larry Laudan pointed to theories that predicted well and posited things we now say do not exist: caloric, phlogiston, the luminiferous ether. Why expect our current entities to fare better? Bas van Fraassen's constructive empiricism says science aims only at theories that fit the observable phenomena; belief in unobservables is optional. Ian Hacking offered a practical way out: when physicists spray electrons onto a target to change its charge, they are using electrons as tools, and what you can use that reliably is real, whatever your theory says about it. As someone who spends his days manipulating things he cannot see, I find that persuasive.

What is an explanation?

Finally: when have you explained something rather than merely described it? Carl Hempel said an explanation derives the event from laws plus initial conditions. The classic objection is the flagpole. From the pole's height and the sun's angle you can derive the shadow's length, which is a fine explanation. But from the shadow's length and the angle you can derive the pole's height, and the shadow plainly does not explain the pole. Explanation has a direction that logic alone does not capture. Most philosophers think causation supplies it; van Fraassen argued that explanations answer why-questions relative to a contrast, which depends partly on our interests.

Three habits for a beginner

Philosophy of science does not hand you a rule that sorts claims into science and nonsense. What it gives you is better habits. Ask what would count against a claim, and be wary of claims that nothing could count against. When a prediction fails, ask which assumption is being protected and whether the protection predicts anything new. And distinguish confidence in what a theory lets you do from confidence in everything it says about the unseen.

My friend, for the record, still reads her horoscope. She says it is a form of play. I have decided that is an honest answer, which is more than its authors usually manage.

A good first book is Alan Chalmers's What Is This Thing Called Science? Then read Kuhn's Structure, which is shorter and more readable than its reputation.