The idea was to match people not based solely on similarities (unless that’s what a participant values in a relationship), but on complex compatibility questions.Each person would fill out a detailed survey, and the algorithm would compare their responses to everyone else’s, using a learned compatibility model to assign a “compatibility score.” It then made the best one-to-one pairings possible — giving each person the best match it could — while also doing the same for everyone else.
What if they gave people one match based on core values, rather than many matches based on interests (which can change) or physical attraction (which can fade)?
They’ve run the experiment two years in a row, and last year, 7,600 students participated: 4,600 at Stanford, or just over half the undergraduate population, and 3,000 at Oxford, which the creators chose as a second location because Sterling-Angus had studied abroad there.
“There were videos on Snapchat of people freaking out in their freshman dorms, just screaming,” Sterling-Angus said.
But it’s unclear if the project can scale beyond the bubble of elite college campuses, or if the algorithm, now operating among college students, contains the magic key to a stable marriage.
The idea was hatched during an economics class on market design and matching algorithms in fall 2017.