I think what you should be doing as a dating app — much more so if a government one — is some kind of RL based on length of actual marriage.

At least, you should have some algorithm estimating marriage length / likelihood of divorce based on your (non-app) statistics, and then use that as one factor for your in-app calculations.

Just to call out, you're implicitly assuming that strictly lower divorce rates is the correct end-goal. This may not be what you really want, eg you might find many abusive pairings where the weaker partner is too scared to leave. In the style of "The optimal amount of fraud is non-zero" [1], I'll claim that the optimal amount of divorce is non-zero.

[1] https://www.bitsaboutmoney.com/archive/optimal-amount-of-fra...