Researchers develop ‘troll-detector’

Academics build model to zero in on anti-social commenters after just five posts

Tags: EntertainmentUSA
  • E-Mail
Researchers develop ‘troll-detector’
By  Stephen McBride Published  April 20, 2015

US academics have discovered common characteristics among website commenters who go on to be labelled trolls, The Guardian reported.

Researchers at Cornell and Stanford studied posts on three sites (CNN, Breitbart and IGN, all of which use posting platform Disqus) over 18 months.

More than 35m posts from almost 2m users were examined. Researchers also had access to deleted or reported comments. Some 50,000 users were barred from sites during the period under investigation.

The researchers found that future trolls had a tendency to write comments that were less readable than their better-behaved peers. Trolls’ comments also would depart from common lingo used by the posting community and use fewer words associated with positive emotions.

“We find that such users tend to concentrate their efforts in a small number of threads, are more likely to post irrelevantly, and are more successful at garnering responses from other users,” the researchers wrote in a report called “Antisocial Behaviour in Online Discussion Communities”.

“Studying the evolution of these users from the moment they join a community up to when they get banned, we find that not only do they write worse than other users over time, but they also become increasingly less tolerated by the community.”

Fledgling trolls also react badly to strict moderation, the researchers found, noting that negative behaviour tended to escalate following action by site moderators that was seen as overly harsh.

From the data, the academics were able to build a statistical model that could predict with 80% accuracy if a given user would eventually be banned, by examining their first five comments. The authors hope the model can be used to target antisocial behaviour earlier and more effectively.

Add a Comment

Your display name This field is mandatory

Your e-mail address This field is mandatory (Your e-mail address won't be published)

Security code