The forum they used as a source of data was a media outlet during a presidential election. What, precisely, is the community that they are imagining the comments section consists of? It is by definition contentious. Furthermore, antisocial behavior cannot possibly be defined as something you do as a group. It's in the name! Political vitriol is a team sport.<p>Trollish behavior is also mutually exclusive with anger, for the simple reason that we already have a pretty good understanding for the cause of peoples behavior when they are angry. Anger. The fact that discussions degenerate when people are angry is, presumably, WHY the study looks to explain trollish behavior.<p>My understanding of the methodology, when you leave out the CNN data, was as follows (Please correct me if I'm wrong, because my opinion is scathing).<p>First they garnered a set of participants with no overarching demographic, presumably to ensure that the results were as broadly applicable as possible, but I think failing to remember that online communities tend to be much more homogenous.<p>They then found a political article on reddit and seeded a new comment section for some with some reddit comments on the original article. They made 32 copies of this comment section, randomized the order of the comments and assigned 20 people to each.<p>Before they were allowed to comment, they were given a quiz designed to manipulate their mood by asking hard or easy questions, and then lying about how well they did relative to the median. They then justified the use of this particular piece of psychological warfare as having an effect which correlated with positive and negative moods, which are then interpreted as the causal element in the behavior that follows.<p>Immediately before commenting, the users are instructed that they are "testing a new voting system", which is just, excuse my language, the shittiest instruction I have ever seen participants given in a study.<p>Then they measured something other than the stated definition of trolling, and declared their results to be significant.<p>I won't comment on the stated definition of trolling except to say that a strong case can be made that these stanford/cornell researchers participated in behavior which was both dishonest and disruptive, with the effect of pissing at least me off. Does that not qualify?