Declared vs Revealed Preferences
Declared preferences are what people say they want when you ask them. Revealed preferences are what they actually do, measured by behaviour rather than by self-report. The two frequently disagree, and Nir Eyal's design instruction in Hooked (2014) follows from taking that gap seriously: build for the cat videos people actually watch, not the cinema-quality home movies they say they want to make. The connection this site draws further is that watching behaviour more closely does not, on its own, close the gap, because revealed behaviour tracks wanting rather than liking, and those are different systems.
What it is
Ask someone what they want from a product and the answer tends to be aspirational, flattering and only loosely tethered to what they will actually do once the product is in front of them. This is not because people are lying. Most people are genuinely bad at introspecting their own motives, and the answer that comes out under questioning is shaped by what sounds reasonable to say out loud, to a stranger, about oneself. Declared preference is that self-reported answer. Revealed preference is the pattern that shows up in behaviour instead, session length, click paths, what gets opened again and what gets abandoned, measured independently of anything the person claims about their own intentions.
Nir Eyal's design instruction in Hooked draws directly on this distinction. He credits the design researcher Erika Hall with its cleanest statement: people say they want to make cinema-quality home movies, and then they sit and watch cat videos. Designing for the declared version, the home-movie ambition, gets the product wrong almost by default, because it is answering a question about self-image rather than a question about what will actually get used. Evan Williams, who founded Blogger, Twitter and Medium in succession, generalised the same instinct into a product principle of his own: find a human desire that predates the technology by decades or centuries, then use new technology to strip steps out of satisfying it, rather than trying to invent a new desire from scratch. Jack Dorsey ties Twitter's design to the same habit of attention from a different angle, writing detailed narrative sketches of a hypothetical user's actual context and mood, the story they are already living, rather than the story that user would tell about themselves if you interviewed them.
Where this gets more complicated than "trust behaviour, not surveys" is the connection to Wanting vs Liking, Kent Berridge's dissociation between the brain system that drives pursuit and the separate system that produces enjoyment. Revealed preference, in the form of a click, a reopen, a returned session, is generated by the wanting system. It is not generated by, and does not reliably track, the liking system, the part of the same architecture responsible for whether the experience was actually enjoyed. That matters because it means watching behaviour more carefully is not a route to discovering what people like. Every behavioural trace available to a product team is, by construction, downstream of wanting rather than liking, so no amount of additional behavioural data closes that particular gap. The only way to learn about liking is still to ask, which puts a version of the declared-preference question back on the table even after behavioural data has answered the revealed-preference one.