Wanting and liking are not the same thing
May 27, 2026 · 読了7分
A familiar observation
You scroll the menu for twenty minutes. Add, remove, swap toppings, compare prices, open two more places just to be sure. The feeling is unmistakable: keyed up, focused, slightly restless.
Then the food arrives. You eat it. It is fine.
The gap between keyed up and fine is something a lot of people notice and rarely say out loud, because saying it sounds odd. If it was only fine, what were those twenty minutes.
Two systems, not one
Kent Berridge, a neuroscientist at the University of Michigan, spent years with his colleagues on a question that sounds like it should not need asking: is wanting something the same thing as liking it.
The answer coming out of the animal work was no. Berridge and colleagues showed that you can intervene in the dopamine system and change how hard an animal pursues a reward, while the reactions that indicate pleasure at actually tasting that reward stay roughly where they were. The two can come apart.
He gave them names: wanting and liking.
- Wanting is the drive toward. It is loud, forward-pointing, and tied to the dopamine system.
- Liking is the actual pleasure during consumption. It is quieter, shorter, and runs on different circuitry.
To be clear about what this is: a description of a research programme, not a clinical claim about you. Reasoning from animal studies to human behaviour always needs care, and Berridge himself has been careful about it. But the two-system framing accounts for a lot that a one-system model cannot.
Why menus are so strong
If wanting and liking come apart, which one does product design aim at?
The louder one. The one that pulls people forward.
Delivery apps are extremely refined machines for wanting. Photos shot at the best angle under the best light. Promo countdowns. A line telling you portions are running out. The price stepping up as you add toppings, and that step itself producing a small hit. The entire flow from opening the app to hitting order is a chain of tiny actions, each with immediate feedback.
The liking half is not something they control. Food goes cold in transit. Packaging leaks. The dish does not look like the photo, because the photo was never the dish. But all of that happens after the money has moved, which puts it outside what gets optimised.
Reward prediction error
There is a related idea that usually comes up alongside: reward prediction error. In short, the dopamine system responds most strongly not to a reward itself but to the gap between the reward received and the reward that was expected.
This matches ordinary experience. The usual dish, exactly as expected, produces very little. The new place, the thing you have never tried, the part you cannot predict — that is where the signal is.
So when you scroll a menu at midnight, you are not looking for calories. You are looking for the unpredicted part. An effectively infinite catalogue always has more unpredicted part left, which is why scrolling can outlast eating by a wide margin.
Where this leads
If most of the pleasant part of ordering food sits in choosing rather than eating, there is a fairly simple possibility available: keep the choosing, drop the rest.
That is precisely what Table for None does. You get the whole menu. You get the price jumping when you add an egg. You get the countdown and the map with a rider crawling across it. What you do not get is a meal, and what you do not lose is money.
This is not true for everyone and not true at every hour. If you are genuinely hungry this is a pointless exercise and you should go cook something — every dish in the catalogue carries a recipe under twenty minutes for exactly that case.
But if what you are chasing is the wanting rather than the liking, the wanting does not need a real order to run on.
The best part of a midnight delivery usually happens before the delivery.
Knowing that does not fix anything, and should not be sold as fixing anything. But it makes skipping an order feel less like giving something up and more like subtracting the right part.