What you're guessing from

Ten chapters of this book have come down to the same thing, which is that the guess is wrong in one direction. You expect less pleasure and more awkwardness than there is, and you judge your wording while they hear your warmth. It's worth asking why a guess that's wrong so consistently doesn't correct itself, because you've been making it for years and it hasn't budged.

Nicholas Epley and four colleagues, Michael Kardas, Xuan Zhao, Stav Atir and Juliana Schroeder, published a review in 2022 that pulls the compliment findings together with a lot of similar ones, on talking to strangers, on expressing thanks, on small kindnesses, and gives the pattern a name, undersociality. People are less social than would be good for them and for the people around them, because they expect their overtures to be received worse than they are. The review suggests three reasons the guess stays wrong, and I'll put them in plain terms.

The first they call differential construal. You and the recipient aren't looking at the same thing. You're looking at your compliment as an act you're performing, and grading the performance. They're looking at it as news about what you think. Two people, one sentence, two completely different objects.

The second is uncertain responsiveness. You don't know how they'll respond, and when people don't know, they assume the worse case and act to avoid it. So you don't say it, and because you didn't, you never find out that the worse case wasn't coming.

The third is asymmetric learning, and I think it's the one that keeps the drawer shut. You learn from what you do. When you say the good thing and it lands well, you get a data point. When you keep it in, you get nothing, no landing, no evidence, no correction. Every unsaid compliment leaves the guess exactly where it was. And the rare time it goes badly, the brush-off or the blank look, that's the data point you remember, because it confirmed the squirm. It's like an archer who never sees where the arrows go and only hears about the one that missed.

That's the machine, and I'd say it's why understanding matters here rather than resolve. You can't correct a guess you never test. The only way the guess gets better is by saying the thing and watching the landing, and that's what the second half of this book is for.