What Would Change Your Mind?

A friend tells you he's certain his business partner is skimming money. You ask how he knows. He says he can just feel it. So you ask the only question that matters: what would you have to see to believe you were wrong? He goes quiet. Then, slowly: "Nothing, really." That is not a man with strong evidence. That is a man with a strong feeling and no exit. A belief you can never abandon isn't a conclusion. It's a tattoo.

There is one question worth asking before any decision that matters: what would change my mind? Not as a rhetorical flourish. As a real spec, written down, with an answer. If you can't name the evidence that would flip you, you don't have a belief, you have a mood. The whole skill of thinking under uncertainty fits inside that question, because answering it forces all four moves at once, in order, before your ego has time to dig in.

Decide the test before you see the result

The trick is timing. You name your threshold before the verdict comes in, not after. This sounds obvious and is almost never done, because a threshold set after the result is just your conclusion wearing a lab coat. Say you're sure a new hire is a disaster. Fine. Before week two, write the bar: "If she ships two clean projects and the team stops complaining, I was wrong." Now you've built a door you can actually walk through. Set that bar after week two, and you'll quietly slide it to wherever you already stand.

A threshold set after the result is just your conclusion wearing a lab coat.

That is move one and move two folded together. First, name the belief in plain words — "this hire is a disaster" — so it's a claim and not a fog. Then check the base rate, which just means: of all the new hires who felt rocky in week one, how many actually turned out to be disasters? Not the loud, memorable ones. All of them. If most shaky starts turn fine by month three, your gut is arguing against the odds — and the odds get a vote.

Strong evidence versus the loud kind

Now move three: weigh the evidence. The test for any new fact is brutally simple. Would I expect to see this if my belief were true — and would I also expect to see it if my belief were false? Strong evidence is likely under one story and unlikely under the other. Weak evidence shows up either way and just feels persuasive because it arrived with volume.

Take the skimming partner. "He seemed nervous at lunch" is weak: nervous people exist under both stories — guilty and innocent-but-stressed-about-rent. It splits nothing. A wire transfer to an account nobody can explain is strong, because that's common if he's stealing and genuinely rare if he isn't. One of these survives the test. The other is a vibe — and most of what wins arguments at dinner is the vibe, delivered confidently, at length.

There's a quiet tell for weak evidence: it confirms whatever you already think and never points the other way. If every fact you collect happens to support your side, you're not gathering evidence, you're shopping. Real evidence occasionally embarrasses you. The friend who only ever finds reasons his partner is guilty isn't investigating. He's decorating a verdict he reached at the start.

Update, don't collapse

Then the part everyone fumbles. The strong evidence arrives, and the urge is to swing the whole way — from "he's fine" to "he's a criminal" in one lurch. That's collapse, not updating. One strong fact moves the needle; it rarely pins it. A single unexplained transfer should take you from "no concerns" to "genuinely worried, time to look at the books" — not to a confession speech. You moved. You didn't teleport. The size of the move should match the size of the evidence, and almost nothing in real life is one clean step from doubt to done.

What makes the question so efficient is that it does the calibration for you. Naming your threshold in advance is just deciding, while you're still honest, how much evidence a real verdict costs. When the proof shows up, you're not negotiating with yourself in the heat of it — you're checking a receipt you wrote earlier. The work gets done before the emotion arrives, which is the only time the work is ever clean.

So design the test, then read the result. Not the other way around. A belief that costs nothing to hold and nothing to drop is worth roughly what you paid for it. Your friend, by the way, never opened the books. He preferred the feeling. It was, in fairness, a very strong feeling. It was just never going to be evidence — and somewhere underneath he knew it, which is exactly why he made sure it could never be tested.