Google-Logo-AH-1

Google Backed Research Intended To Identify Trolling Behavior Accurate 8 Out Of 10 Times

April 23, 2015 - Written By John Anon

As strange as it may be, it is difficult to fully define what ‘trolling’ is or who a ‘troll’ might be. Unlike their more mythical namesake, internet trolls are more often defined by their behaviors than any physical or meaningful characteristic. Therefore, some define trolls as those who say things for the sake of saying them. There is no reasoning or logic behind the comments and instead, are more designed to upset, annoy or instigate further interaction. Others, define trolls as those simply seeking attention and those who will say anything to obtain that attention. The one consensus which seems to be in tune across definitions is that trolls are ‘anti-social’. They do not positively impact on social based constructs like forums or group discussions, but instead negatively impact on people and conversations.

Well, Google recently backed a study which looked to identify the traits of trolling behaviors. By attempting to identify these traits, the idea was to be able to predict which users are likely to be trollers, long before their trolling tendencies became evident. The study was a joint research project between researchers at Stanford University and Cornell university and spanned a period of 18 months. Over that time frame, more than forty million posts from 1.7 million users, across three major sites (CNN, Breitbart and IGM) were observed and analysed to try and identify the cursors underlying trolling behaviour.

The results were primarily defined by looking at those users who eventually became banned (presumably for trolling) and comparing their behaviors to those who were not banned (presumably the non-trollers). The results highly suggested that those who do troll can be predicted in as little as ten posts, and the success rate of these predictions was noted in the 80-percent region. Some of the tell-tale signs of the future trollers, were found to be a combination of concentrated posting (meaning that they were far less active across posts and seemed to much often focus on specific posts), coupled with a significant increase in the number of posts, posted. For instance, the researchers found that ‘future trolls’ posted as many as 264 times in a similar time period to those (who were defined as non-trolls) would post only 22 times. A tenfold increase in posting. As such, their behaviour was far more active while also being far more selective. Interestingly, in spite of this concentration, the future trolls were also noted as being able to elicit more responses and replies from other members, suggesting that trolls are good at engaging other members in “fruitless, time-consuming discussions”. More to the point, that the larger non-trolling community maybe more often than not, fuelling the issue of trolling.