(Kimihiro Hoshin0/AFP)

Here is a bit of news that should surprise no one: Despite myriad incremental efforts to make the Internet more woman-friendly, misogyny is still a scourge on social media.

That finding comes to us courtesy of the think tank Demos, which published some updated figures on Twitter abuse today as part of a British anti-harassment campaign. According to the think tank, Twitter users — male and female — sent 200,000 aggressive tweets using the words “slut” and “whore” during a three-week period in late April and early May. (That works out to roughly 6.6 abusive tweets per minute, or 9,500 per day.) Those tweets were directed to 80,000 people, including Hillary Clinton, Michelle Obama and Beyonce.

But maybe what’s most interesting about this report is not the new numbers themselves — instead, it’s the way Demos went about finding them. As the think tank explained in a news release, it scraped the Twitter firehose for instances of the words slut and whore, and then used a proprietary natural-language processing program called Method52 to determine which instances were jokes/endearments and which were actually abusive. In other words, Demos deployed an automated tool to accurately flag abuse without a user having to manually report it.

That’s significant, because victims of online harassment — misogynistic and otherwise — have been asking Twitter to build that kind of automated tool for years: a means to identify abuse, and escalate it to moderators, without victims assuming the burden of flagging and reporting tweets themselves.

Method52 was developed by the University of Sussex and the Center for the Analysis of Social Media in 2012; Twitter’s version — “a product feature to help us identify suspected abusive Tweets and limit their reach” — has been in testing since April 2015. And the site’s automatic quality filter is still not available to everyone.

All of which leaves us wondering: What gives, Twitter? There’s still a whole lot of ladies getting called sluts on your platform.

Liked that? Try these: