Skip to main content

New troll catching system works surprisingly well

Are you partial to a bit of trolling? Well, shame on you first of all, but also be warned that there could soon be much more robust systems in place to detect and then deal with troll posters – at least if the latest developments from US researchers are anything to go by.

Researchers from Stanford and Cornell Universities have crafted an algorithm which can pinpoint trolls quite swiftly and with a fair degree of accuracy, Wired reports.

The research was funded by Google, and examined the behaviour of trolls over a period of 18 months (across sites including CNN and IGN) to enable key identifiers to be determined – and the system was honed so it only needed to observe between five to ten posts to predict whether a specific user would need to be hit with the ban stick.

Those key indicators included, unsurprisingly, posts that more frequently showed poor spelling and grammar, and posts which degraded in quality over time (with less thought being put into following troll posts). They also contained far more profanities, and other negative language designed to stir other users up.

The system does throw up some false positives, though, and one in five were actually errantly identified as trolls, the authors of the paper observed. They report stated: “While we present effective mechanisms for identifying and potentially weeding antisocial users out of a community, taking extreme action against small infractions can exacerbate antisocial behaviour (e.g., unfairness can cause users to write worse).

“Though average classifier precision is relatively high (0.80), one in five users identified as antisocial are nonetheless misclassified. Whereas trading off overall performance for higher precision and have a human moderator approve any bans is one way to avoid incorrectly blocking innocent users, a better response may instead involve giving antisocial users a chance to redeem themselves.”

Twitter recently made a move to help deal with trolls, via a new feature called a ‘quality filter’ which is designed to remove any malicious tweets from a user’s account.