The Facebook Core Data Science Team’s 2012 experiment on “massive scale emotional contagion” was one of the group’s first high-profile endeavors in the realm of the human sciences to appear in a major journal (PNAS, 2014). As it turned out, it was also the last. Although the team went on to publish widely in other fields, it has not continued to make available its work in the human sciences. This paper explores how language-processing software—explicitly trained to harvest narratives of trauma—was employed in the study to quantify emotional valence. In an atmosphere of rapid social and technological change around 2012, researchers applied processing software and computational analysis to alter the algorithms delivering manipulated News Feeds to almost 700,000 unwitting Facebook users. The experiment formed a watershed in creating coercive conditions around social media exposure. I examine how emotional contagion—a pre-existing concept in psychological studies but one that researchers calculated and computed in new ways in 2012—was operationalized and with what consequences.