WORLD NEWS
Feynman’s 50-year-old restaurant memo revealed when to stop ordering a new menu
Researchers have deciphered physicist Richard Feynman’s restaurant choice notes from nearly 50 years ago and proven that his solution was mathematically optimal.
It is not an all-purpose formula for choosing a restaurant, but is the result of calculating the balance between exploration and repetition under simplified experimental conditions in which the quality of all restaurants is fixed.
What was the problem?
Let’s say you’re eating dinner several nights at a travel destination. You can try a new restaurant every day or go back to the best place you’ve ever been. The goal is not to find the best one, but to achieve the greatest sum of dining satisfaction for the entire trip.
What was Feynman’s answer?
In the beginning, try new places, but once you find a restaurant that exceeds a certain level of quality, repeat that place for the remaining days. The fewer days left, the lower the bar required for new selections. By restoring Memo’s equation, the researchers proved that this strategy is optimal for a specific quality distribution.
How did they choose 2,520 people?
In a pre-registered online experiment, participants chose a virtual restaurant with a hidden score. Instead of calculating an exact optimal curve, people used the simple rule of lowering the bar uniformly as time goes by. This method required a little more searching than the optimal solution, but the performance was almost similar.
Can I use it the same way in real life?
The taste of a real restaurant changes every day, and price, distance, company, and mood also affect your choice. The research model assumes that once you visit, you will know exactly the quality and can visit again. Therefore, this should not be misunderstood as a rule of thumb to stop at a certain number of restaurants on a given day.
why is it meaningful
The same conflict of exploration and exploitation appears in homes, parking spots, jobs and online recommendations. It is meaningful in that it provides a standard for comparing how well simple rules used by humans perform instead of complex optimal calculations.