From time to time a really bad post on a social network gets a lot of attention. Say a head of state falsely accuses a journalist of murder
, or suggests that mail-in voting is illegal — those would be pretty bad posts, I think, and most people working inside and outside of the social network could probably agree on that. In my experience, though, average people and tech people tend to think very differently about what to do about a post like that. Today I want to talk about why.
When a tech company employee sees a really bad post, they are just as likely to be offended as the next person. And if they work on the company’s policy team, or as a moderator, they will look to the company’s terms of services. Has a rule been broken? Which one? Is it a clear-cut violation, or can the post be viewed multiple ways?
If a post is deeply offensive but not covered by an existing rule, the company may write a new one. As it does, employees will try to write the rule narrowly, so as to rule in the maximum amount of speech, while ruling out only the worst. They will try to articulate the rule clearly, so that it can be understood in every language by an army of low-paid moderators. (And who may be developing post-traumatic stress syndrome and related conditions.)
Put another way, when an average person sees a really bad post, their instinct is to react with anger. And when a tech person sees a really bad post, their instinct is to react practically.
All of that context feels necessary to understand two Twitter debates playing out today: one over what Twitter ought to do about the fact that President Trump keeps tweeting without evidence that one of the few high-profile Republicans who regularly speaks out about him, the onetime congressman and current MSNBC host Joe Scarborough, may be implicated in the 2001 death of a former staffer. And one over what to do about the president’s war on voting by absentee ballot
As to the former: In fact, according to the medical examiner, former Scarborough aide Lori Klausutis died of a blood clot
.) Now her widow is petitioning Twitter CEO Jack Dorsey to remove Trump’s tweets suggesting there may have been foul play. John Wagener wrote up the day’s events in the Washington Post
With no evidence, Trump has continued to push a conspiracy theory that Scarborough, while a member of Congress, had an affair with his married staffer and that he may have killed her — a theory that has been debunked by news organizations including The Washington Post and that Timothy Klausutis called a “vicious lie” in his letter to Dorsey.
On Tuesday morning, Trump went on Twitter again
to advocate the “opening of a Cold Case against Psycho Joe Scarborough,” which he said was “not a Donald Trump original thought.”
“So many unanswered & obvious questions, but I won’t bring them up now!” Trump added. “Law enforcement eventually will?
If you believe social networks are obligated to remove posts that are indecent, it’s clear why you would want these tweets to come down. The president is inflicting an emotional injury on an innocent, bereaved man for political gain. (Trump has historically benefitted from falsely suggesting his Republican opponents are murderers, as Jonathan Chait notes here
But if your job is to write or enforce policy at a tech company, your next steps are far less clear. Consider the facts. Did Trump say definitively that Scarborough committed murder? He didn’t — “maybe or maybe not
,” he tweeted this morning. Did Trump incite violence against Scarborough, directly or indirectly? (Twitter has promised
to hide such tweets behind a warning label, but it has never done so.) I don’t think so, and while encouraging law enforcement to investigate the case arguably represents an abuse of presidential power, our nation’s founders invested the responsibility for reining in a wayward chief executive not with private companies but with the other two branches of government.
Let’s make it more complex: Scarborough is a public figure — a former congressman, no less. Traditionally social networks have tolerated much more indecency when it comes to average people wanting to yell at the rich and powerful, and when it comes to the rich and powerful yelling at one another. And when two of those figures are engaged in political discourse — the kind of discourse that the First Amendment, which informs so many of the principles of tech company speech policies, sought to protect above all else — a tech policy person would probably want to give that speech the widest possible latitude.
I spent the day talking with former Twitter employees who worked on speech and policy issues. For the most part, they thought Trump’s Scarborough tweets should stay up. For one, the tweets don’t violate existing policy. And two, they believe you can’t design a policy that bans these tweets that doesn’t also massively chill speech across the platform. As one former employee put it to me, “If speculation about unproven crime is not allowed, I have bad news for anyone who wants to tweet about a true crime podcast.”
Now, it’s possible for me to imagine a time when Twitter would have to take action against these tweets. There was a time when Alex Jones’ tweets and videos about the Sandy Hook school shooting also fell into the realm of “speculating about true crime,” even though his conspiracy theories were almost certainly promoted in bad faith. But then Jones’ fans began stalking and harassing families of the murder victims, in some cases threatening to kill them
. Eventually Jones was removed from most of the big social platforms
If Trump continues to promote the lie about Scarborough, we can assume some of his followers will take matters into their own hands. It’s been barely a year since one of those followers was sentenced to 20 years in prison for mailing 16 pipe bombs
to people he perceived to be Trump’s enemies. If something similar happens as a result of the Scarborough tweets, Twitter will face criticism for failing to act. It’s a terrible position for the company to be in.
But mostly it’s just a terrible thing for the president to do. And in a democracy we have remedies for bad behavior that go well beyond asking a tech company to de-platform a politician. You can speak your mind, you can march in the streets, and you can vote. That’s why, for most problems of political speech, my preferred solution is more speech, in the form of more votes.
Which brings us to the day’s surprising conclusion: Twitter’s decision to label, for the first time, some of Trump’s tweets as potentially misleading. Makena Kelly has the story in The Verge
On Tuesday, Twitter labeled two tweets from President Donald Trump making false statements about mail-in voting as “potentially misleading.” It’s the first time the platform has fact-checked the president.
The label was imposed on two tweets Trump posted Tuesday morning falsely claiming that “mail-in ballots will be anything less than substantially fraudulent” and would result in “a rigged election.” The tweets focused primarily on California’s efforts to expand mail-in voting due to the novel coronavirus pandemic. On Sunday, the Republican National Committee sued California Gov. Gavin Newsom
over the state’s moves to expand mail-in voting.
According to a Twitter spokesperson, the tweets “contain potentially misleading information about voting processes and have been labeled to provide additional context around mail-in ballots.” When a user sees the tweets from Trump, a link from Twitter is attached to them that says “Get the facts about mail-in ballots.” The link leads to a collection of tweets and news articles debunking the president’s statements.
This story is surprising for several reasons. It involves Twitter, a company notoriously prone to inaction, making a decisive move against its most powerful individual user. It ensures a long stretch of partisan mud-wrangling over which future tweets from which other politicians deserve similar treatment — and over whether one side or another is being punished disproportionately. And it puts Twitter prominently in the position it has long sought to avoid — “the arbiter of truth
,” chiming in when the president lies to say that no, actually, it’s legal to vote by absentee ballot.
And yet at the same time, Twitter’s decision was rooted in principle. In January Twitter began allowing users to flag tweets that contain misleading information about how to vote
. Today it applied that policy, fairly and with relative precision. Some have criticized the design and wording of the actual label — “Get the facts about mail-in ballots” doesn’t exactly scream “the president is lying about this.” But it still feels like a step forward, and not a small one.
Social networks that reach global scale will always suffer from really bad posts, some of them posted by their most prominent users. And it’s precisely because those platforms have become so important to political speech that I would rather decisions about what stays up and what comes down not be dictated by the whims of unelected, unaccountable founders.
Twitter’s decision to leave up some of Trump’s awful tweets and label others as misleading won’t fully satisfy anyone. But in my view this is a case where the company has made some hard decisions in a relatively judicious way. And anyone who tries to write a better, more consistent policy — one that goes beyond “this is indecent, take it down” — will find that it’s much harder than it looks.