Falsifiability: The Test for Real Information

Someone says, "The government is corrupt." Another replies, "No, it isn't." They go back and forth for an hour, and neither moves an inch. Why? Because neither claim was ever built in a way that evidence could settle.

Falsifiability is what helps us notice and take account of exactly that.

What It Is

Falsifiability is a simple test: a claim only carries weight if you can say what evidence would prove it wrong. If no evidence could ever count against it, it is not telling you anything about the world. It is closer to a preference, stated in the form of a fact.

Notice what the test does not ask. It does not ask whether the claim is true. It asks whether the claim is the kind of thing that could be shown false. A claim has to risk being wrong before it is worth anything.

Where It Comes From

The idea is most closely tied to Karl Popper, the twentieth-century philosopher of science. Popper was bothered by a question: what separates real science from things that merely sound scientific?

His answer was that a good theory sticks its neck out. Einstein's general relativity, for example, predicted that starlight would bend as it passed the sun, and astronomers could check that during an eclipse. Had the starlight not bent, the theory would have been in trouble. It took the risk, and it survived.

Compare that with a theory built so that every possible outcome counts as confirmation. Whatever happens, it was predicted. Popper's point was that such a theory looks strong but is empty. If nothing can ever count against it, it explains nothing.

The Test in Practice

Take the claim from the opening.

"The government is corrupt." As stated, there is nothing to check. Corrupt in which office, doing what, compared to what? Any example you bring can be waved away, and any example the other person brings can be too.

"This ministry awarded a tender to a company owned by the minister's relative, above the next best bid." This is a different kind of statement. Tender records exist. Ownership records exist. Bid amounts exist. Someone can look, and the claim can turn out wrong. Because it can lose, it is worth something.

Notice that the broad claim may still feel true to the person saying it, and the specific examples underneath it might even be true. But the broad statement itself never risks anything, so it never settles anything.

Three Levels of Difficulty

Falsifiability is not a simple switch between "testable" and "untestable." Real claims come in levels.

Level one: clean and testable. "The site is slow because of the database." You can test this directly. Replace the database call with a fake instant response and see whether the slowness disappears. If it does not, your explanation was wrong. This is the easy case, and the habit to build is asking what single change would prove an explanation wrong, and whether anyone has actually tried it.

Level two: testable, but only if you isolate the variable. "Prices went up because of the new tax." This sounds testable, but prices move for many reasons at once: fuel costs, exchange rates, supply problems. Seeing prices rise after the tax does not show the tax caused it. To test it properly, you would compare against a similar place with no such tax, or trace how the tax actually reaches the shelf price. This is where most political, workplace and personal claims quietly fail. They are not untestable. Nobody has separated the variable that would confirm or kill the claim from everything else moving at the same time. So before asking whether the claim is true, ask what else could explain what you are seeing.

Level three: human and interior claims. "She has been distant since Tuesday." You cannot run a controlled experiment on someone's inner state, but that does not mean there is nothing to go on. You can check what is observably different: fewer messages, shorter replies, less eye contact. The honest move is to separate what you observed from the story you built on top of it, and to hold the story more loosely than the observation. The evidence here is partial, not absent, and feelings deserve to be taken seriously for exactly that reason.

Where It Shows Up

Politics and public debate. Slogans like "we are finally serious about the economy" cannot lose an argument, because they never risk a checkable prediction. A statement like "this policy will lower the price of unga within a year" can be checked, and that is why it is more honest, even if it turns out wrong.

Everyday comparisons. "This phone is better" or "this team is better" is often several claims bundled together: camera versus battery, attack versus defence. Two people can argue for an hour while defending different claims. Ask what, specifically, is better, and what result would show otherwise.

Work. "This feature just feels heavy" cannot be settled. "This feature adds 400 milliseconds to page load" can.

Self-talk. "I'm just bad at meetings" is a verdict. "I lost the thread because I had not read the agenda" is a testable explanation you can do something about.

Falsifiability and Its Neighbours

It works closely with other models on this site. Confirmation bias is the habit of looking only for evidence that supports what you already think. Falsifiability is the discipline of asking what evidence would count against it. First principles thinking asks what you actually know to be true as opposed to merely assumed, and a falsifiable claim is one you could actually test against that.

Where It Runs Out

Be honest about the limits.

Falsifiability is a test of how a claim is built, not a test of how much a claim matters. Some of the most important parts of life, such as values, meaning and loyalty, are not testable claims, and they do not need to be. The mistake is not holding such beliefs. The mistake is presenting a belief as a factual claim, or treating an untestable statement as though it had settled a factual argument.

It also does not make a claim true. A claim can be falsifiable and still false, and a claim you cannot test today may become testable tomorrow when better tools arrive. The test tells you whether a claim is in a state to be argued about with evidence. It does not decide the argument for you.

Questions to Ask

Before you agree with a claim or argue against it, ask three things:

1. What specifically is being claimed? Define the terms until the claim is concrete.
2. What would prove it wrong? Name the evidence, and check whether anyone has looked for it.
3. What wasn't controlled for? List what else could explain the same result.

The Actual Lesson

Much of what gets argued about, at work, online and at home, was never a claim in the first place. It was a feeling, a loyalty or a slogan, dressed up to sound like a conclusion. Falsifiability does not stop you from having feelings or loyalties. It simply helps you notice which one you are dealing with before you spend an hour arguing about it.

So the next time someone makes a confident statement, try the quiet question: what would it take to show this wrong? The answer, or the silence that follows, tells you most of what you need to know.