Google AI hopes to offer fresh Perspective on toxic comment sections

23 Feb 20174 Shares

Share on FacebookTweet about this on TwitterShare on LinkedInShare on Google+Pin on PinterestShare on RedditEmail this to someone

The New York Times website front page. Image: Goran Bogicevic/Shutterstock

Share on FacebookTweet about this on TwitterShare on LinkedInShare on Google+Pin on PinterestShare on RedditEmail this to someone

Google is turning to its AI developers in a bid to clean up toxic comment sections on websites, with a new platform called Perspective.

While comment sections on websites started out as a commendable effort to bring a bit of discourse to a story or opinion held by an author, the past few years have put enormous strain on its moderators.

Despite their entire job being to police comments for outright and unnecessary abuse against other users and the author, the sheer number being generated either by humans or programmed bots is becoming overwhelming.

Ranking of toxicity

According to previous surveys, 72pc of American internet users claim to have witnessed or experienced online harassment, affecting roughly 140m people in the country.

To help combat this, Google’s internet safety incubator Jigsaw and counter-abuse technology team have developed a new technology that will employ AI to give them a helping hand.

Called Perspective, the open source API uses machine learning to predict the impact a comment might have on a conversation.

This will be shown as a rating scale, with the words most associated with abuse – such as swear words or insults – ranking as the highest in toxicity.

Perspective will then alert the human moderator of a high toxicity rating, or automatically hide that comment for users behind a warning message.

Perspective

An example of Perspective at work. Image: Google/Jigsaw

Will expand into other languages

It will also attempt to get the commenter to think twice about what they’re saying, with the AI revealing how toxic it thinks the phrase is.

Interestingly, although Jigsaw said that the API will be left open at the publishers’ discretion, it will allow readers to sort comments by toxicity themselves.

Jigsaw has been working with The New York Times, using Perspective to help it sift through the 11,000 comments it receives each day, which is currently done manually.

The next step, according to Jigsaw’s president Jared Cohen, is to expand Perspective into languages other than English, and help the AI to determine when a comment is off-topic.

“In the long run, Perspective is about more than just improving comments. We hope we can help improve conversations online,” Cohen said.

The New York Times website front page. Image: Goran Bogicevic/Shutterstock

Colm Gorey is a journalist with Siliconrepublic.com

editorial@siliconrepublic.com