clock menu more-arrow no yes

This web app will warn you if you are a Twitter troll

An intern at Intel created it as part of the “Hack Harassment” effort.

Intel intern Alexei Bastidas
Ina Fried for Recode

Rather than spend the summer developing new chips, Alexei Bastidas spent his internship at Intel teaching a computer how to spot harassment on the internet.

The result is a web app, currently in testing, that tells people just how intimidating they are on Twitter, offering both a numerical rating as well as example tweets that could be seen as harassing.

It’s the first technological component to emerge from Hack Harassment, an Intel-led effort to see if there are high-tech ways to make internet communities safer. (Disclosure: Recode and Vox Media, which owns this site, are among Intel’s partners in the campaign.)

Although the aim of the service is to identify harassing speech, Bastidas said he trained the algorithm using a wide range of internet content from the darkest parts of Reddit to mainstream news sited like the New York Times and the Wall Street Journal, as well as Twitter itself.

“We’ve joked around that we’re trying to give our algorithm a liberal arts education,” Bastidas said. The goal is to release the tool publicly (as well as the source code behind it) over the next couple of months.

The move comes as Twitter itself is under fire for not doing enough to prevent harassment and a growing number of publishers including NPR have eliminated their comments sections. Just this past week, Twitter introduced new tools designed to at least better mask harassing tweets.

“I think technology is where this is happening,” Intel senior VP Doug Fisher told Recode after a Hack Harassment panel at this week’s Intel Developer Forum. “Technology has a role to play to figure out ways to help in all this.”

Bastidas ran his algorithm on my tweets to show how it worked. It initially choked on my nearly 43,000 posts.

An analysis of a subset of my more recent tweets rated me as 99 percent nice. I’m thinking they need to tighten their algorithm a bit. Just ask Mike Isaac.

Ina Fried for Recode

This article originally appeared on Recode.net.

Sign up for the newsletter Sign up for The Weeds

Get our essential policy newsletter delivered Fridays.