Chrome extension from Google wants to filter out toxic comments

The comments sections of websites bring out the best and worst of the internet --  engaging insights are often found right next to enraging insults.

Unless website moderators keep watchful eyes, comments sections can devolve into chaos flooded with toxic posts that drive away conversations. Moderating comment sections has become so difficult, websites like The Atlantic, Vice and NPR have decided to shut them down entirely.

"If publishers and websites really want to get people engaging with their content, they need to figure out the comments section, and right now they're still very toxic," said Jennifer Grygiel, a social media professor at Syracuse University.

Enter Tune, announced Tuesday by Alphabet's Jigsaw. The new Chrome extension automatically hides comments that its Perspective AI considers toxic.

The tool will only work for comments on Reddit, Twitter, Facebook, YouTube and Disqus. Its like a volume control for internet comments, CJ Adams, a product manager at Jigsaw, said in a blog post.

Jigsaw first rolled out Perspective in 2017, offering AI to comments sections of platforms like The New York Times and Wikipedia. Perspective learned to flag negative comments from thousands of people labeling millions of posts as spam, harassment or obscene content.

"Most of us spend more time reading online comments than writing or moderating them," Adams said in the post. "As we read, a single toxic post can make us give up on a discussion completely and miss out on reading valuable thoughts buried underneath the shouting."

Google implemented Perspective for YouTube, allowing channel owners to block out comments that the AI found was toxic. The tool was originally created to help moderators keep their comment sections under control, and now the Chrome extension allows anyone to control what they see online.

"This abuse is an industry-wide issue and it should be all hands on deck across the industry," Grygiel said.

You can toggle how many comments you want to see -- from complete silence in "Zen Mode" to essentially viewing everything online without a filter.

You can also adjust it by site, in case you only want to block toxic comments on Facebook but want to see every post on Reddit, for example. You're also able to filter it based on certain terms, if you're specifically looking to block negative comments with threats, sexually explicit attacks and insults, Google said.

The "Filter Mix" setting is still "very experimental," and will have a significant amount of false positives.

Perspective isn't perfect, and Adams said that it could still miss toxic comments or accidentally block comments that are perfectly fine. When the AI was first launched, it mistakenly considered trash talk over sports to be online harassment.

Jigsaw said Perspective is constantly learning, and it's asking for Tune users to leave feedback on the extension to improve its algorithm. It won't be automatically collecting users' data, as Google wants to avoid people purposely tricking its AI.

"When people manually leave feedback and false positives, we use that data to help train and improve the models," a Google spokesperson said.

Jigsaw's hope is that Tune shows platforms a new way to handle trolls and harassment campaigns, without a need for comment moderations.

While Tune may block out negative content for its users, it's not blocking it from the internet -- which means harassment will ensue, you just might not see it.

But if a comment calls me an idiot, and no one was there to read it, was I really insulted?