Experiment № 064 · The story
Who Asks First
Why I picked it
Sixty-three experiments of things that flow, grow, erode, sync and search — and not one of them about people with opinions about each other. Gale and Shapley's 1962 paper is nine pages of the American Mathematical Monthly with no computer anywhere near it, and it settles a question you'd swear needs one: given two sides who rank each other, can you always pair everybody so that no two people would rather ditch their partners for each other? Yes, always, and here is the procedure. What sold me is that the procedure is embarrassingly simple — ask, hold on provisionally, trade up, repeat — and that it was already quietly running the American medical residency match a decade before anyone proved it worked. Shapley shared the 2012 Nobel in economics with Alvin Roth for what came of it; Gale had died in 2008.
What I enjoy about it
Two counters running in opposite directions, which you can watch on the little rank strips. The asking side starts at its first choice and can only ever slide further down its list. The answering side starts with whoever turned up and can only ever trade up, because it never lets go of anyone until something better knocks. Nobody in the whole picture ever checks whether the matching is stable — there is no global test anywhere in the code — and yet the counter for pairs who'd both walk falls to zero and stays there. I also love the cruelty of the middle of a run: a couple can look settled for twenty asks and then break in one frame because someone better finally got round to knocking. Turn the shared-taste dial to 100% and the drama vanishes entirely — when both sides agree on who's desirable there is only one stable matching, and it doesn't matter who speaks.
Something you might take away
Stable is not the same as fair. This algorithm doesn't find a stable matching, it finds a very particular one: the best stable outcome that exists for every single person on the asking side, and simultaneously the worst for every single person on the answering side. Flip the toggle and the same preferences, the same rules and the same definition of success hand the advantage straight to the other team. That's not a quirk — the residency match ran hospital-proposing for decades and switched to applicant-proposing in 1998, once it was clear which way the thumb was on the scale. So when someone describes a process as neutral because everyone's preferences are taken into account, the question worth asking is a much smaller one: who gets to propose, and who is only allowed to answer?