In a candid blog post today, Twitter gave an update on its ongoing effort to tamp down on hateful speech, threats, and other abusive behaviors on its platform.
It says that in the first three months of this year, it suspended 100,000 account holders who creating new accounts following an initial suspension (a 45% year-over-year increase) and that it flagged three times more abusive accounts within 24 hours after a report. It also revealed that now, thanks to machine learning and other automated techniques, 38% of abusive content on Twitter is “surfaced proactively” for its team of human reviewers, and that it’s observed 16% fewer abuse reports after an interaction from an account the reporter doesn’t follow.
“The same technology we use to track spam, platform manipulation and other rule violations is helping us flag abusive Tweets to our team for review,” wrote Twitter vice president Donald Hicks and director of product management David Gasca. “With our focus on reviewing this type of content, we’ve also expanded our teams in key areas and geographies so we can stay ahead and work quickly to keep people safe. Reports give us valuable context and a strong signal that we should review content, but we’ve needed to do more and though still early on, this work is showing promise.”
Twitter says that furthermore, it’s responding 60% faster to appeals requests on average (in part thanks to its recently introduced in-app appeal process), and that it’s removing 2.5 times more private information from reports.
“People who don’t feel safe on Twitter shouldn’t be burdened to report abuse to us. Previously, we only reviewed potentially abusive Tweets if they were reported to us,” wrote Hicks and Gasca. “We know that’s not acceptable, so earlier this year we made it a priority to take a proactive approach to abuse in addition to relying on people’s reports.”
In the coming weeks, the company says it intends to “improve [its] technology” to help review content that breaks its rules more quickly and make it easier for Twitter users to share specifics when reporting. Additionally, it says it’ll add more notices within Twitter’s suite of apps for clarity and enable people to hide replies to tweets (as previously announced), and says it’ll update its rules so they’re “shorter,” “simpler,” and “easier to understand.”
“[W]e’ve made strides in creating a healthier service,” said Hicks and Gasca. “Keeping people safe on Twitter remains our top priority, and we have more changes coming to help us work toward that goal.