Bayes' Rule: How to Change Your Mind Without Losing It

Here is a small, annoying truth: most people update their beliefs at exactly two speeds — not at all, or all the way. A reassuring text arrives and they're suddenly certain everything is fine. A weird symptom shows up and they're certain it's a catastrophe. Bayes' rule is the boring third option between those two, and it happens to be the correct one.

Strip away the math and Bayes' rule is a recipe for changing your mind by the right amount. You start with a belief. New evidence shows up. You adjust. The whole trick lives in the right amount — not zero, not everything, but proportional to how telling the evidence actually is.

Three words, no equations

Your prior is where you start — your belief before the new evidence walks in. Wake up with a headache and your prior isn't a coin flip; it's whatever headaches usually mean for you, which is almost always nothing. That starting point is the base rate, the plain background frequency of a thing. Begin there, not at the scary end of the scale.

Then the evidence arrives, and not all evidence is equal. A vague hint barely moves you. A strong, specific sign moves you a lot. The question to ask is sneaky but simple: how much more likely is this clue if I'm right than if I'm wrong? A clue that shows up either way tells you almost nothing.

Your posterior is just your updated belief — prior plus evidence, blended. Not a fresh start, not a clean slate. The old belief is still in there, pulled in a new direction. Posterior simply means after. That's the entire vocabulary, and you now know more Bayesian statistics than most people who use the word at parties.

Watch it move

This is what the machine below is for. Drag the slider to set your starting belief, then click a piece of evidence and watch where it lands. A weak hint nudges. A strong sign shoves. Counter-evidence drags you back. Notice the thing it refuses to do: it never snaps straight to 100%, no matter how many times you click.

Refusing to hit 100% is the healthy part, not a bug. Certainty would mean no future evidence could ever change your mind — you've welded the door shut. Bayes keeps a sliver open on purpose. A bad feeling is data. It is not a conclusion, and the slider quietly enforces that distinction.

So next time something startles you, run the four moves: name the belief, check the base rate, weigh how telling the evidence really is, then update without collapsing. Your brain is loud. That doesn't mean it's right. Move the slider a notch, not to the wall, and you'll be wrong less often than the people sprinting between certainties.