Twitter Is Expanding the Signals It Uses to Detect Bad Behavior

Cleaning up the site, one tweet at a time

Twitter is changing how it displays tweets that it suspects might be abusive. Twitter
Headshot of Marty Swant

In its quest to further decrease the abusive content that litters the site, Twitter said it has begun using “behavioral signals” like tweets from unconfirmed emails and people signing up for multiple accounts at once.

Today, the San Francisco-based company announced updates to its algorithm that will “improve the health of the public conversation” on Twitter by identifying tweets from trolls even before they’re reported by users. In addition to email addresses and related accounts, Twitter will also identify users who repeatedly tweet and mention accounts that don’t follow them.

The enhanced detection tools are the latest update on Twitter’s announcement in March that it wants to change everything from the verification system to how it crowdsources ideas from outside individuals and companies. (While Twitter often pitches itself as a place to keep up with the news, it’s also continued to be a place that’s full of offensive people talking about the news.)

While the company would not say outright the tools are related to bots, anyone who has mentioned a celebrity and been replied to and retweeted by dozens of random accounts will likely understand why Twitter is doing what it’s doing. The updates might also help Twitter’s efforts to decrease bots on the platform such as the ones with ties to Russian accounts that have been used to harass journalists, celebrities and other users in coordinated attacks.

“Some troll-like behavior is fun, good and humorous,” according to a blog post published written by Twitter vp of trust and safety Del Harvey and director of product management and health David Gasca. “What we’re talking about today are troll-like behaviors that distort and detract from the public conversation on Twitter, particularly in communal areas like conversations and search. Some of these accounts and Tweets violate our policies, and, in those cases, we take action on them. Others don’t but are behaving in ways that distort the conversation.”

Because many of these types of tweets don’t directly violate Twitter’s content rules, the company can’t always remove them altogether, according to the post. And while such tweets will remain on Twitter, they won’t show up right away unless a user clicks on “show more replies” or chooses to see everything within search. The company said early global testing has led to a 4 percent decrease in abuse reports from search and an 8 percent decrease in abuse reports from conversations.

Twitter wasn’t the only social media platform to reveal new measures today for cleaning up its user-generated content. This morning, Facebook sought to address its own transparency issues by disclosing how it’s found and removed millions of pieces of offensive content so far this year along with more than 800 million fake accounts.

Earlier this month, as Twitter announced plans to debut 30 new shows, CEO Jack Dorsey made a rare appearance onstage in New York City to yet again pitch his platform. And while he didn’t take any questions, he again mentioned how Twitter is focused on conversation.

“We love seeing the world come together in one place on our platform and talk about what’s happening,” Dorsey said.

@martyswant Marty Swant is a former technology staff writer for Adweek.