Trolls, those rude people who pollute Twitter or news websites with their insulting comments, may have their days numbered. Justin Cheng and several colleagues at Stanford University have developed an algorithm to unmask them at the first turn .
Researchers have analyzed the behavior of users blocked for their antisocial behavior on three pages with considerable traffic: CNN.com, the political news portal Breibart.com and the specialized video game website IGN.com.
When comparing their messages with those of other non-offensive posts, they found that their readability (measured with a test called the Automated Readability Index) was poorer and, in addition, the quality worsened over time.
Cheng and his colleagues believe that the behavior and (bad) style of trolls is so predictable that they can be automatically unmasked, thanks to their algorithm, between the fifth and the tenth post.
However, the inventors admit that they must refine their system, as it can give "false positives" and troll a harmless user . For now, human moderators are still essential, but in the future tools like this can help preserve good manners on the internet, always within the limits of freedom of expression.
