Complex Algorithm Could Help Identify Online Trolls Before They Strike

Whether you are playing an online game, spending some time in a chat room or just commenting on a blog post, there’s always the risk of running into an online troll. It’s true that their comments are sometimes childish and easy to brush off, but when these people turn to threats and heavy insults, things can get serious really fast. Multiple large websites, including the Huffington Post, spend a considerable amount of money on comment moderation, but not all websites can afford to pay for such a clean job. It would be a lot easier if these individuals could be identified and banned before they could hit “send”, and the good news is that recent research shows that this might be possible in the near future.

A team of researchers over at Stanford and Cornell examined patterns of behavior displayed by about one in 40 users that visit the IGN, Breitbart and CNN websites. They focused on what exactly makes a troll be a troll, and they realized that most of them are unwilling to use language that’s appropriate for an online community. They also took note of their inclination towards swearing and the volume of contributions that they make to a debate. By using these patterns to create an algorithm, researchers believe that they can identify trolls with an accuracy of 80% by analyzing just five of their online posts. If they manage to perfect the process, internet trolls might soon become extinct, at least on some of the world’s largest websites.

What do you think about this new research?

Thank you Economist for providing us with this information.

Image courtesy of Hackinformer.