What, then, is an interview actually for? The question sounds naive and the answers people give are incompatible, which is why the genre is such a mess.
The commercial answer is promotion, and it accounts for the overwhelming majority of interviews ever conducted. The late-night talk show as Johnny Carson perfected it was a machine with an explicit function: a celebrity has a film opening, the celebrity appears on the couch, an anecdote is told which has been tested in the pre-interview, the film is mentioned, everybody profits. There is nothing dishonest about it as long as nobody pretends otherwise. The seven-minute junket interview is the same transaction stripped of charm, and the reason those interviews are unbearable is that the interviewer is contractually obliged to pretend to be curious.
The accountability answer is Paxman's, and it is real but narrow. When someone with power is evading a question, making the evasion visible in public is a genuine service, and the fact that no information is transmitted is not the point — the demonstration is the point. What has gone wrong with the genre is that the adversarial posture escaped its proper habitat and became a general style, so that people with no power are subjected to prosecutorial questioning about things nobody is accountable for, and everyone learns to answer like a minister.
The intimacy answer is the modern one, and it has produced the dominant format of the last decade: the very long conversation, three hours, no adversarial framing, in which a listener comes to feel they know the guest. Length is not the problem; length is a genuine gift, and some of the best interviews ever recorded are long. The problem is length without preparation, which does not produce information — it produces self-presentation at greater volume. Given three hours and no resistance, an intelligent person will construct an extremely persuasive account of themselves, and the listener will mistake the duration of exposure for depth of access. It is possible to spend three hours with someone and learn nothing they did not intend you to learn.
Terry Gross is the counter-case and the finest practitioner of the accountable-intimate mode. Her interviews are prepared in enormous detail, they are gentle, and they extract things that the guests plainly did not arrive intending to say. She has spoken about the strange advantage of working remotely — she in Philadelphia, the guest in a studio elsewhere — because the absence of physical presence removes the social obligation to be nice, on both sides, and lets a hard question be asked without the room turning cold. That is a technical insight of the first order, and it is the same insight as Cowen's about context-shifting, arrived at from a completely different direction.
Larry King is the perfect inversion, and the honest one. He did not prepare, on principle. His argument was that reading the guest's book made him an expert and he did not want to be an expert; he wanted to be the audience, asking what the audience would ask. There is a real theory there and it produced a career of forty years. It also produced thousands of hours in which nothing was learned by anyone, and it is the exact opposite of everything Cowen does.
So: promotion, accountability, intimacy, and — the rarest — extraction. Cowen's answer is unambiguous. The interview exists to obtain information that is not otherwise obtainable, from a person who has it, for an audience that wants it. Everything that does not serve extraction is stripped out: the biography, the warm-up, the interviewer's views, the audience's comfort, the guest's comfort, the arc.
It is not a warm format. It was never trying to be. And it has produced, across a decade, a body of primary material about how unusually productive people actually work that has no real precedent, assembled by a man whose entire method depends on his willingness to spend three weeks reading someone's collected papers in order to ask them eleven good questions.
from Reading Tyler Cowen: Growth, Taste, Talent, Markets, AI, and the Art of Having Too Many Interests (2026)