Gale-Shapley, used in this service, has the property that one side ("proposing" side) is guaranteed their most advantageous stable match while the other side is guaranteed their worst possible stable match.
I've never heard this discussed or mentioned as a drawback because it's Nobel Prize winning and "stable" is a shorthand for "fair". Even though there's many unique stable matches that each offer participants different utilities even if individual participants don't have incentives to change their pairing.
In practice, the more powerful party seems to be the proposing side, e.g. hospitals are the "proposing side" when using Gale-Shapeley to match w/ med school students and get their ideal students.
I wonder if anyone who works on dating apps knows if men or women are the preferred side? And is there any research on stable matching that ensures equity (or ideally arbitrary distribution of utility) between the two groups in their stable matches?
This metric/logic equation does not exist in any dating company, also I highly suspect that anyone working there has hear of the term.
Here is why:
A dating sites primary KPI are mainly "new signups per day" and "subscriptions signed per day". For a lot of reasons listed above,dating sites in general _cant_ever_ work for the majority of people.(yes,there is this one neighbour who is happily married, they exist, sure, but this is not the majority)
Integrating something like Gale-Shapley would require your company to focus on othe matching part and integrate it somehow - but there is no reason for you to do so, because if you keep the primary KPI in the correct range, you are printing money, regardless if your matching algo is scientic sophisticated or not. So why bother?
Apart from that - also this algo cant check for chemical compatibility and friendship compatibility.
jjmarr · · focus · HN ↗
I've never heard this discussed or mentioned as a drawback because it's Nobel Prize winning and "stable" is a shorthand for "fair". Even though there's many unique stable matches that each offer participants different utilities even if individual participants don't have incentives to change their pairing.
In practice, the more powerful party seems to be the proposing side, e.g. hospitals are the "proposing side" when using Gale-Shapeley to match w/ med school students and get their ideal students.
I wonder if anyone who works on dating apps knows if men or women are the preferred side? And is there any research on stable matching that ensures equity (or ideally arbitrary distribution of utility) between the two groups in their stable matches?
KellyCriterion · · focus · HN ↗
Here is why:
A dating sites primary KPI are mainly "new signups per day" and "subscriptions signed per day". For a lot of reasons listed above,dating sites in general _cant_ever_ work for the majority of people.(yes,there is this one neighbour who is happily married, they exist, sure, but this is not the majority)
Integrating something like Gale-Shapley would require your company to focus on othe matching part and integrate it somehow - but there is no reason for you to do so, because if you keep the primary KPI in the correct range, you are printing money, regardless if your matching algo is scientic sophisticated or not. So why bother?
Apart from that - also this algo cant check for chemical compatibility and friendship compatibility.