The Washington PostDemocracy Dies in Darkness

Twitter’s internal numbers show it’s getting less awful at fighting abuse

(Eric Thayer/Reuters)

Twitter on Thursday offered some top-line stats in a company blog post about the results of its latest anti-abuse efforts — showing how its approach is working as social media firms struggle with policing content on their networks.

Online abuse is a problem across all social networks. But Twitter in particular has the reputation for being a hotbed for online abuse and for having a lot of trouble containing it. A recent BuzzFeed article outlined several troubling cases in which Twitter seemed blind to abusive language or agonizingly slow to respond.

The statistics offered Thursday indicate there's been some progress in Twitter's fight against abuse — but didn't offer hard numbers, apart from Ed Ho, the company's general manager of consumer product and engineering, disclosing that Twitter suspends or limits “thousands more abusive accounts each day.” Overall, according to Thursday's post, Twitter takes action on 10 times the amount of abuse per day than it did at the same time last year.

That could be read in a couple of ways. It is a sign that Twitter is dealing with more abusive content, or it shows abuse is rampant, and Twitter was handling only a fraction of what was out there before.

In 2016, it formed a“Trust and Safety Council,” with members from several anti-abuse, anti-harassment and online safety groups. And in January, the company publicly said it's redoubling its efforts across the company.

On the product side, Twitter has also updated its policies to give users tools to avoid abuse directed at them. Users can choose, for example, not to be notified when people outside their network mention their usernames — a common tactic for abusers. Since introducing this policy, the company has seen 40 percent fewer blocks in that scenario, which it takes as a sign that people aren't seeing as much abuse on the site.

And looking at its persistent whack-a-mole problem, where abusers simply make new accounts once their old ones have been suspended, Twitter said that it has doubled the number of these accounts it has removed.

These are small moves. But Twitter's efforts speak to shifting views toward regulating speech on all social media networks, said Stephen Balkam, chief executive of the nonprofit Family Online Safety Institute. Balkam, a member of Twitter's Trust and Safety Council, said the network has “come a long way” from a time early in the company's history when it seemed to view free expression as being far more important than community safety. But, he said, there is a way to let people speak without threatening anyone's safety, and it's worth figuring out how to get there.

“Our democratic values are messy, and more difficult and more challenging — but we would prefer that to censorship,” Balkam said.

Del Harvey, Twitter's head of trust and safety, said some policies seem to change how people act on Twitter. The company now sometimes puts abusive users' accounts into a sort of penalty box for a limited time, which keeps them from sending tweets or retweeting.

Penalized accounts, Twitter said, generate 25 percent fewer abuse reports going forward. And the majority of users who are slapped with limited functionality — 65 percent — don't end up as repeat offenders, the company said.

That’s not because they never come back,” Harvey said. “They are still tweeting.” And that, she believes, is a good thing — and a sign that maybe Twitter's tools are changing behavior, at least a little, without having to lose users.

Still, Harvey and Ho acknowledge that these numbers still aren't that good.

“This is going to be a long-haul thing,” Harvey said. “Our work here will never, ever be done.”

Correction: An earlier version of this story misstated when the Trust and Safety Council was formed.