In this study, published in the Proceedings of the National Academy of Sciences on June 17, 2014, Adam Kramer of Facebook with Jamie Guillory and Jeffrey Hancock of Cornell ran a large experiment on the Facebook News Feed. For one week in January 2012, the feeds of 689,003 users were altered: for some users the algorithm reduced the share of posts containing positive words, and for others it reduced the share containing negative words. The researchers then measured the emotional tone of those users’ own subsequent posts.
The finding was that emotional states are contagious at scale and through text alone. When positive content was reduced in a person’s feed, that person produced slightly fewer positive words and more negative words in their own posts; when negative content was reduced, the opposite happened. The effects were very small per person but, multiplied across hundreds of thousands of users, the authors argued they were meaningful. The study was notable because it showed contagion without face-to-face interaction and without users being aware of it.
The paper became a landmark in research ethics as much as in social science. It drew sharp criticism because the users had not given informed consent to having their feeds manipulated for an experiment. PNAS published an Editorial Expression of Concern noting that the work may not have followed standard principles of informed consent and the opportunity to opt out, while declining to retract the paper.
Why business readers should care: this is the clearest documented demonstration that the choice of what a feed shows can shift users’ emotional state, and the ensuing controversy set much of the agenda for how platforms and researchers think about experimenting on users.