Last night, developer Victoria Fierce had some harsh words for Vice President Mike Pence. Incensed over the Trump administration’s recent rollback of transgender protections, Fierce let loose on Pence. “Fuck you,” she tweeted. “I gotta piss, and you’re putting me — an American — in danger of assault by your white supremacist brothers.”
Almost immediately, she got the notification. Twitter had detected “potentially abusive activity” on her account, and put her in temporary timeout as a result. For the next 12 hours, only followers could see her tweets — which meant she wouldn’t be able to lobby the Vice President.
It’s part of a broader shift in the way Twitter deals with abuse — one that’s drawn fire from many of the same communities calling for stronger enforcement. In some cases, the account-lock could even be triggered by quote-tweeting a profanity-laced tweet, since the resulting tweet would include both profanity and the source’s handle.
Still, the abrupt notification caught Fierce by surprise. “It was just one tweet,” Fierce told The Verge, “and certainly not the first time I’ve told an elected official to fuck off.”
Twitter says the account throttling is tied to changes that began rolling out last week designed to diminish the reach of abusive accounts. When Twitter determines that an account is being abusive — using criteria that the company has declined to describe in any detail — it hides the account’s tweets from anyone who isn’t following it. The restrictions last for a set period of time, typically 12 hours for a first offense.
For the average recipient of abuse, that’s good news: when a troll quote-tweets you with an insult, you simply won’t see it. But the limits are raising new questions in light of the political speech that is often hosted on Twitter. The average person has an expectation that their account will not receive insulting tweets on a regular basis. But what about, say, the president of the United States?
All the high-profile incidents have involved verified accounts, but a spokeswoman told The Verge that Twitter’s content filters do not currently distinguish between types of accounts. A single tweet will not trigger its filters — you can quote-tweet Vice President Mike Pence and call him a dick if you want to — but a “pattern” of abuse will. The company has declined to offer more details about what constitute a pattern or triggers its content filters, saying that trolls will use the information to game the system. But in The Verge’s own tests, single instances of abuse were not enough to trigger the block.
The content filters are continually evolving, the spokeswoman said, and Twitter is “thinking a lot about” the use of strong language in relation to political speech.
The shift comes after sustained criticism of the way Twitter handles abuse, which Twitter has often struggled to respond to. Earlier this year, the company removed notifications telling users that they’ve been added to a list — ostensibly a response to abusive or hostile list names — only to quickly roll back the change after users complained it might actually enable abuse.
For Fierce, the new lockdowns fall victim to a similar problem. “Politically, I think this sends a message that Twitter considers all forms of abuse equal,” she says. “To their systems, a white supremacist calling for shooting a person of color is just as bad as an angry Latinx renter telling their city’s rent board to fuck off.”