Why one good go doesn't fix it

Suppose you say it. Tonight, in the kitchen, you tell your partner the call was handled well, and it lands, and you can see it land. The guess was wrong and now you know. Job done, you'd think. The drawer's open and the habit's gone.

I'd like that to be true and I don't think it is, and there's a recent finding that says why. Stav Atir and Nicholas Epley published three experiments in 2026 on what happens to people's expectations after a conversation with a stranger. Conversations with strangers, like compliments, go better than people expect. And having one did move people's expectations, made them more optimistic and more accurate, especially about talking to the same person again. But the correction was fleeting. Within one or two weeks, expectations had drifted back to the same pessimistic baseline as people who'd never had the conversation. The good experience didn't stick. The guess reasserted itself, like a tan.

That study is about talking to strangers, not about compliments, and I'm extending it. I think the extension holds because the guess in both cases comes from the same three mechanisms in chapter 13, and nothing about a compliment makes it easier to learn from. But it's my extension, and you should hold it as one.

Here's what follows, if it does hold. One good go proves the point for about a week. After that the squirm is back at full strength, telling you it'd be awkward, as though the kitchen never happened. You'll have the memory of it landing and you'll still not do it, because the memory has stopped being a belief and gone back to being a story about one Tuesday. That's not weakness. It's how the correction wears off in the experiments too.

So the fix can't be a single brave act. It has to be a standing arrangement that doesn't ask the guess's permission, one that runs whether or not you currently believe it'll be fine. That's why chapter 22 gives you a rule rather than a pep talk. One a day, said regardless, and the guess gets tested every day instead of once. The guess doesn't need convincing. It needs outvoting, daily, by evidence it can't keep up with.

That's also why this book is longer on what stops you than on what to do. What to do takes ten seconds. What stops you has been reinstating itself every fortnight for years, and it'll go on doing that until saying the good thing is something you do rather than something you once did.