As a former FB employee who worked on ranking, I just want to point out how absolutely tiny the effect size is here when we think about normal human emotional e

As a former FB employee who worked on ranking, I just want to point out how absolutely tiny the effect size is here when we think about normal human emotional experience. On a per person basis, this experiment had an effect that is comparable to other potential changes like slightly altering the padding on the "like" button, or showing 1 extra news post per day, or sending an extra notification once per month. Facebook didn't make people sad. It made a population post things with slightly more negative words, only significant when measured across hundreds of thousands of users.