http://www.businessinsider.com/facebook-study-on-language-and-personality-2013-10
Well, yeah, they would get more accurate results observing what people do in there natural environment (in this case what they would have posted any way) than they would with a more traditional, "answer these questions no one cares about" personality study. They'd get even more accurate results if those being studied didn't know they were being studied. And even better ones if the subjects didn't know that studies existed.