In the United States at large, your right to free speech is protected by democratic institutions. Online, however, who gets to say what is based on the judgment of unelected corporations. What you say, and what you see, depends on judgments made by the guardians of the social media “commons”.
Two of the guardians, Facebook’s chief operating officer, Sheryl Sandberg, and Jack Dorsey, chief executive of Twitter, testified before Congress on Wednesday. Elected officials expressed a number of doubts about the companies’ effectiveness in monitoring speech. The Senate hearing was devoted in large part to concerns over manipulation by foreign agents, the presence of malicious bots, and the safeguarding of private user data. The House hearing, at which only Dorsey was present, dwelled more on conservatives’ perception that there is a leftwing bias on Twitter, and that conservative accounts are being “shadowbanned”.
Both Dorsey and Sandberg had prepared well for the hearings, mastering the art of satisfying the congressional inquisitors without actually saying anything substantive beyond promises like “trust is the cornerstone of our business”. Replying respectfully to the bluster of congresspeople, they promised to do better, to be more transparent. What they didn’t do, however, was actually disclose how their censorship decisions work. And that should leave people worried.
Facebook and Twitter have immense power over public discourse. When they decide content isn’t fit public consumption, it can disappear forever, as if into a black hole. Yet both Sandberg and Dorsey were vague when asked to describe the processes by which speech is “disappeared”.
Dorsey said that Twitter’s “singular objective is to increase the health of public conversation” and it is “now removing over 200% more accounts for violating our policies”, but didn’t offer many specifics about how the judgment calls are made.
Sandberg encouragingly said that in determining whether a post is “fake news”, “we don’t think we should be the arbiters of what’s true and what’s false”. But what they do instead is turn to “third-party fact-checkers”. If the fact-checkers flag a post as false, Facebook will “dramatically decrease the distribution” and “show related articles so people can see alternative facts”.
We don’t know much more, after several hours of congressional testimony, about what causes a Twitter account to be suspended or a Facebook post to be removed. What we heard isn’t reassuring. Outsourcing “truth policing” to a third-party fact checker can only work if the fact-checkers themselves have sound and trusted judgment. Sandberg and Dorsey’s promises to combat bullying and harassment were laudable, but the crucial question is always: how does the company decide what constitutes harassment?
Content-filtering algorithms aren’t reliable decision-makers: they have not even proven themselves “able to distinguish between child nudity and a historical atrocity” and Facebook even accidentally flagged the Declaration of Independence as “hate speech” (to be fair, it contains the phrase “merciless Indian savages”.)
But as much as we might not want Jack Dorsey and some secret algorithms to have unilateral decision-making power over the online commons, the hearings showed why government regulation may be even worse. House Republicans seemed to want to force Twitter to make sure conservatives were treated fairly, and the farcical spectacle of the hearing (at one point, a Congressman began impersonating an auctioneer in order to drown out a protester), did not give confidence that more intensive government oversight would be wise.
There is, fortunately, an existing model of an online institution that is policed by neither corporate nor government power: Wikipedia. Everyone’s favorite free encyclopedia (and homework aide) is a genuine democratic community, where policies about what gets said are decided communally and are completely transparent. Though Wikipedia has been criticized, especially in its early years, for the presence of errors, it hasn’t suffered from the kind of “fake news” scandals that have plagued other platforms. It has been called the “good cop” of the internet, and it’s so reliable that YouTube and Facebook have turned to it to reliably rebut untruths.
Wikipedia’s reliability is due to its democratic model: people trust it because it achieves consensus through a transparent process. Because people can see what’s going on, and participate in decisions about content themselves, there are no mysteries about why certain content appears and other content is removed. Nobody has more access to knowledge of the practices than anybody else, and with an elected “supreme court” and editable core policies, Wikipedia shows how speech regulation can be conducted in a completely transparent and relatively participatory way.
Twitter and Facebook have shown that they shouldn’t be trusted to make tricky decisions about the limits of speech. Congressional representatives make for even worse police officers. There's no perfect way to decide how to deal with lies, hate, and harassment on major platforms. But the best solution is a Wikipedia model for social media: users should define and enforce the terms of service communally, through a messy democratic process.
If Jack Dorsey is really interested in the question he asked Congress: “How do we earn the trust of the people using our service?” there’s an obvious answer: call a “constitutional convention” and let the people themselves see and revise Twitter’s practices.
Nathan Robinson is the editor of Current Affairs
