Your interface should confirm a guess your user already made

Here is a strange thing about perception that took neuroscience a long time to take seriously: your brain is not really a camera pointed at the world. It’s closer to a betting machine. At every instant it has already guessed what it’s about to see, hear, and feel — and what actually reaches your senses is mostly used to check the guess. When the guess is right, almost nothing happens. The signal that travels up your brain, the part that costs attention, is the error — the difference between what was predicted and what arrived.

This is the core of what’s called predictive processing, and once you’ve seen it you can’t unsee it. You don’t notice the weight of your shirt right now because your brain predicted it perfectly. You do notice the one stair that was a half-inch taller than the rest, because your legs bet on a height and the ground disagreed. Surprise isn’t a flavor the brain adds on top of experience. Surprise is the experience your brain spends energy on.

Now put a user in front of your product.

They arrive carrying a thick stack of predictions they didn’t make consciously — a lifetime of other apps, buttons, forms, and gestures. They already “know” that the logo goes home, that the thing in the top right is the menu, that a blue underlined word is a link, that pressing Save will, you know, save. Every one of those is a prediction. And every screen you ship either confirms those predictions or violates them.

When you confirm them, something lovely happens: nothing. The user flows straight to what they came for, and later they’ll describe your product with the highest praise there is — “it just made sense.” What they’re actually reporting is a low prediction error. Their brain spent almost nothing reconciling your interface with the one in their head.

When you violate them, they pay the tax. The Save button that navigates away. The layout that shifts after load so the button they aimed at becomes the one they hit. The word “Projects” that means something different in your app than in the last five they used. None of these are bugs. They all “work.” But each one produces a little spike of prediction error, and that spike has to be resolved by conscious attention — the most expensive resource your user has. Enough spikes and the feeling isn’t “this is broken,” it’s the vaguer, deadlier “this is tiring,” and they leave without ever being able to tell you why.

The build move that falls out of this is almost boringly practical: before you add anything novel, name the prediction it breaks. Not to forbid novelty — a good surprise, the kind that overdelivers on a prediction, is exactly where delight lives. But most surprise in a first-time experience is not delight. It’s noise. So spend your novelty budget on purpose. Keep the plumbing predictable — put things where the rest of the internet already trained your user to look — and reserve the deliberate break for the one moment you actually want them to feel something.

There’s a designer’s rule of thumb that says users spend most of their time on other products, so they expect yours to work like those. That rule always sounded like a concession to me — a vote for boring. Predictive processing reframes it as physics. You are not competing with those other products for attention; you are inheriting the giant model of “how things work” that they built inside your user’s head. You can fight that model and make the user pay to rebuild it, or you can hand them a screen that confirms a guess they already made and let their attention go straight to the part that matters.

The best interfaces feel like they read your mind. They didn’t. They just let your mind finish its own sentence.