Why Trolls Troll

Justin Cheng, a computer science student at Stanford University, gathered 667 participants through an online social media platform to discover what causes trolling behavior. He asked participants to take a quiz, read an article, then engage in a discussion about the article. All participants were given the same article; however some participants were given a discussion that contained troll comments, and others were given discussions with neutral comments.

Based on an analysis of comments on CNN.com, there are two primary factors that influence trolling. The first factor is a person’s mood. Trolling behavior was brought about in an ebb and flow of day/night cycles; the results found that trolling is most frequent late at night and least frequent in the morning, and peaks on Monday. The second factor is the context of a discussion. Discussions that begin with a “troll comment” are twice as likely to be trolled by other participants afterwards, compared to a discussion that doesn’t begin with a troll comment. We feed off of other people’s negativity.

With these conclusions, the experimenters were able to use machine learning algorithms to predict when trolling would occur 80% of the time. Now that we understand what might cause trolling, we can predict when trolling is likely to happen. Therefore, it is easier to train computers to identify trolls, and filter negative content at a more rapid speed.

http://theconversation.com/our-experiments-taught-us-why-people-troll-72798