Feynman’s optimal – A Nobel-winning physicist once turned a Thai restaurant dilemma into a decision-theory puzzle about when to stop chasing novelty. Behavioral scientists have now revisited Richard Feynman’s solution, tested it with 2,520 participants, and found human choices tr
On a late-1970s evening in Glendale, Pasadena’s orbit, Richard Feynman faced the kind of decision that derails dinner plans for almost everyone: ginger chicken—his favorite—on the next order, or a chance at something better by trying something new.
His companion, Ralph Leighton, wondered which way to go. Leighton’s question landed in the real world. but Feynman answered it the way he often did when a problem wouldn’t let him sleep. He grabbed a sheet of paper. scribbled through the dilemma. and worked out a rule: in a simplified model. he calculated a threshold—a number of visits after which the “rational” move would be to stop exploring and always stick to the dish he already knew he liked.
What Feynman turned into a mathematical riddle wasn’t just a quirky story for a physics legend. It sits in decision theory, in a family of problems known as stopping problems. Those are situations where you have to decide whether what you’ve found so far is good enough—or whether it’s worth continuing to search for something better.
Years later. behavioral scientists went back to that Thai restaurant note. some of which had been hidden behind Feynman’s notoriously hard-to-read handwriting. The result is a new study. published in the Proceedings of the National Academy of Sciences on 1 June 2026. that argues Feynman’s solution really was optimal—and that people’s real-world behavior comes very close to matching it.
The chain of ideas starts with that Glendale restaurant visit. when Feynman was a Nobel prizewinning physicist at the California Institute of Technology in Pasadena. Leighton saved the notes. and. in the early 2000s. partially transcribed Feynman’s cursive handwriting in an article he posted online. A decade after that. in 2013. Tom Griffiths at Princeton University in New Jersey transcribed the notes in full for the first time.
At that point, the work might have stopped there—but it didn’t. Brian Christian. a computer scientist and cognitive scientist who is now at the University of California. Berkeley. says the question lay dormant for nearly another decade. In 2021. Christian and Griffiths decided to revisit it. confirming Feynman’s best solution and also solving a generalized version of the problem.
Then came the part that connects the paper to the dinner table. With a third co-author. cognitive psychologist Evan Russek at the City University of New York. Christian’s team asked a straightforward question: if the math is right. do people behave like it when they’re making choices under uncertainty?.
They tested that in an online game with 2,520 participants. Players were asked to imagine visiting a new city for between one and four weeks. Each night. they would choose a restaurant to eat at. earning points for the quality of the restaurant they picked—a number between 1 and 100. Their instructions were to maximize their total number of points.
As the visit got closer to ending, participants became less willing to gamble on novelty. That shift—becoming more conservative as the deadline approached—followed logic similar to Feynman’s optimal formula. The study emphasizes that participants didn’t derive or calculate the mathematical solution themselves. which involves a formula with square roots. Still, their choices closely approximated the strategy.
Behavioural scientist Shoham Choshen-Hillel at the Hebrew University of Jerusalem. reviewing the work. described it as a “super creative article.” She also said the restaurant example stands in for decisions in many settings. including choosing a home to buy. deciding whom to partner up with. and selecting a parking spot.
Christian points to a limit in how faithfully the original model maps onto everyday life. Feynman’s problem doesn’t take boredom into account. The model’s optimal option is to settle on one dish once and for all. In real life. someone might want to keep experimenting on some visits while returning to a trusted favorite on others—say. choosing the same dish every other time and continuing to explore the menu in between.
Even with that caveat. the dinner dilemma remains the same at its core: a tug-of-war between sticking with what feels safe and trying the unknown in hopes of finding something better. In the study’s simplified setting, the pattern shows up clearly. People don’t need the square roots to move toward a stopping rule. They just need an ending to approach.
The research is dated in the calendar—1 June 2026—and anchored to a life story that began in a Thai restaurant. But the feeling it captures is immediate. When you’re staring at a menu you’ve never seen before, you’re not only hungry. You’re making a decision about how much searching you can justify before you commit to the best thing you’ve tasted so far.
Richard Feynman decision theory stopping problems behavioral science dinner choices restaurants cognitive psychology Proceedings of the National Academy of Sciences Shoham Choshen-Hillel Tom Griffiths Brian Christian Evan Russek
So basically if you’ve eaten the chicken you like before, you should just stop ordering new stuff? Idk seems obvious.
Glendale, Pasadena orbit?? Like the restaurant was in space or what. Also 2,520 people sounds made up but okay. I just want my ginger chicken and not a math test.
Wait I thought Feynman was the guy who made jokes and stuff, not behavioral science. But if this is about novelty, then isn’t this just saying people get bored and go back to the same order anyway?
This is kinda wild but also I feel like they’re missing the point. Like the “optimal” number of tries depends on how good the other dishes are, right? If the restaurant is bad, you’d stop sooner. If it’s amazing you keep exploring forever. So the paper is only “optimal” in their fake model not real life. Anyway I’m hungry now.