[DalekBaldwin]

Going further: what if this was broad exploratory study, not meant to test any specific hypothesis? What if they rated posts based on thousands of features, a parameter space large enough to guarantee they would find plenty of correlations (both spurious and substantial), and randomly perturbed each user's feed and monitored how the features of their contributions changed? I would expect that they already do this, and from the very small amount of detail I've seen in articles about the mood factor so far, I wouldn't be surprised if this was just one of the more interesting correlations that fell out of that sort of process.