Algorithms don't predict compatibility

Yes, I do find it is interesting how much faith we have in something by it just being called scientific. But...

About the example of split couples finding themselves 95% or even 100% compatible with their ex's on dating sites: that can be used both ways. The ex's actually DID end up choosing each other and DID have some good times and were, day I say, "compatible" for a set period of time. Plus we're not even sure what drove the divorces in these examples. Did someone cheat? Was there a specific unforeseen event that happened? Heck, the ex's themselves didn't see the split coming in the beginning/match-making part of their relationship.

Have you heard of LikeBright.com? I just found it, but it uses Facebook friendship and "vouching" to play off the "met through a friend" method of finding a relationship.

shudder If I had realized Match.com had bought OkCupid, I would've bailed much sooner. Ironically, days after I shelled out $5 because of OkCupid's seemingly understanding stance on AdBlock, I found my current girlfriend, and shortly thereafter, I closed my account. (I theorize they only match you with good matches once you give them money, especially the $5 once and done deal, where they will never get money from you again.)

Finally, you quote "concrete things like religion, wealth, and education have been associated with relationship happiness." Couldn't they be used in algorithms in a way to meaningful improve the usefulness of matching?