I'm not a machine learning expert, but isn't this an easily solved problem? Similar to TF/IDF, where you mitigate common words by dividing by their overall fre

I'm not a machine learning expert, but isn't this an easily solved problem? Similar to TF/IDF, where you mitigate common words by dividing by their overall frequency, you should be able to divide the weight of any particular "like" by the frequency of likes between the two people. That way a genuine expression of interest by an acquaintance is weighted far higher than a relative or close friend that reflexively likes everything you post.