Does 'Be Curious, Not Judgmental' Work on an AI?
Ted Lasso returns Wednesday, coaching a team he’s not qualified to coach, in a sport he still doesn’t understand, for the fourth season in a row. I respect the consistency.
The scene everyone remembers is the dart game. Ted’s about to win a bet against a man who’s underestimated him for an entire season, and he explains why: people have judged him his whole life, and not one of them was ever curious. If his opponent had asked a single question (“Have you played a lot of darts, Ted?”), he’d have learned Ted played every Sunday for years. Then Ted attributes the lesson to Walt Whitman: “Be curious, not judgmental.”
Whitman never said it. The co-director of the Walt Whitman Archive has checked, which means a man spent part of his career formally confirming that a poet didn’t say a thing on a TV show (academia has its moments). The quote is still right, even with the wrong poet’s name on it.
So does it work on an AI?
The judgmental loop
Watch someone use AI badly and it’s almost always the same movie. They ask a question, get a mediocre answer, and issue a verdict on the entire field: the model is overhyped, the technology isn’t ready. Then they close the tab and go brief their coworkers on it.
That’s a judgment, and judgments end conversations. The mediocre first answer is what you get when the question came with no context and nobody willing to ask a follow-up, which is roughly how my sons used to answer “how was school.”
The curious loop
The people who get real work out of AI run a different loop. When the answer comes back mediocre, they ask what the model assumed and what context it was missing, then they make it show its work and challenge the step that looks wrong.
When one of my agents produces a bad number, I ask it how it got there before I touch anything. Usually the answer is embarrassing in a useful way: it couldn’t see the right system, so it improvised. The model was doing its best with the wrong pipes, and plumbing has fixes (mine are called MCP servers, and I’ve explained that acronym at least four hundred times without losing enthusiasm).
I went to Shimer College, a Great Books school where the entire teaching method was questions. For twenty years that degree looked charmingly impractical. Then the most important skill in technology became interrogating a confident entity that’s sometimes wrong, and the Socratic Method finally has a job market.
Curiosity beats expertise, with one condition
For working with AI, curiosity now beats expertise. The best results I see come from the best interviewers, the people who ask one more question than everyone else was willing to. Deep technical expertise matters less than I expected it to.
The condition is that curiosity without verification is how you end up shipping a hallucination. Ted has Coach Beard, the man who actually reads the books (I have agents that read the books, plus an unread stack on the nightstand that judges me). Your AI needs a Beard too, and it’s you. Stay curious with the model, then check the answer against a source that doesn’t autocomplete.
So, does “be curious, not judgmental” work on an AI? Short answer: yes, better than it works on people, because the AI never gets tired of the questions.
Whitman would agree. He didn’t say that either.