Pages

Friday, May 31, 2019

A Facebook Supreme Court?

New Controversies in Intermediary Liability Law

Anupam Chander

To borrow Churchill’s line about democracy, a Facebook Supreme Court is the worst idea, except for all the others.

In 2018, Mark Zuckerberg introduced the idea of establishing an independent board that would make the most difficult decisions with respect to content. He compared the new body to a Supreme Court. In January 2019, Nick Clegg, Facebook’s Vice-President of Global Affairs and Communications, announced the charter of this independent oversight board. Facebook may have sought to reduce apprehensions of its growing global power by ceding some control to an outside body. However, it was clear that Facebook was borrowing the apparatus, and even the personnel, of government: not only was Facebook implementing a pseudo-judicial body, but Nick Clegg had once served as the Deputy Prime Minister of the United Kingdom.

As Dawn Nunziato has observed, the internet represents the “most powerful forum for expression ever created.” How decisions over content are made on one of the internet’s principal platforms—a platform that connects literally billions of people—is of great importance. Scholars such as Jack Balkin, Tarleton Gillespie, Daphne Keller, Kate Klonick, Thomas Kadri, and Sarah Roberts have powerfully analyzed how internet platforms make decisions on the content that is carried on their sites and the role of intermediaries in free expression today. In my scholarship, I have sought to demonstrate a “First Amendment/?Cyberlaw Dialectic” in which “the First Amendment constituted cyberlaw, and cyberlaw in turn constituted free speech.”

Facebook’s many critics argued that such an oversight board would serve principally as window dressing, or that the introduction of the outside mechanism was merely rearranging the deck chairs on the Titanic. Yet, the Facebook Oversight Board marks a major new experiment. This essay compares the Oversight Board with its alternatives. After all, the Oversight Board must be considered not only on the basis of its flaws—of which there will likely prove to be many—but rather in comparison to its alternatives.

I will consider five alternatives, which I will dub: Mark Decides; Democracy by Likes; Feudal Lords; Judge Robot; and Official Censorship.

Mark Decides

In December 2015, some employees inside Facebook were troubled. A candidate for U.S. president was calling for a ban on immigration for people of a particular religion. Users had flagged the content as hate speech, triggering a review by Facebook’s community-operations team, with its many employees in several offices across the world. In internal messages, some Facebook employees declared that posts constituted hate speech. On its face, a statement targeting a particular religious group for a ban on immigration would seem to violate Facebook’s community guidelines.

Facebook’s head of global policy management, Monika Bickert, explained internally “that the company wouldn’t take down any of Mr. Trump’s posts because it strives to be impartial in the election season.” Facebook explained its decision to the public as follows: “In the weeks ahead, we’re going to begin allowing more items that people find newsworthy, significant, or important to the public interest—even if they might otherwise violate our standards.”

According to the Wall Street Journal, “The decision to allow Mr. Trump’s posts went all the way to Facebook Inc. Chief Executive Mark Zuckerberg, who ruled in December that it would be inappropriate to censor the candidate.” Zuckerberg is the ultimate arbiter of what stayed up or what came down on Facebook.

A central difficulty of Facebook’s current model is that it places enormous power in the hands of Facebook’s leadership and to those employees to whom the company leaders choose to delegate this power. When Facebook deleted a Norwegian government minister’s posting of the famous photograph of the naked Vietnamese girl fleeing a napalm attack, the Norwegian government complained of censorship. Facebook reversed its decision, even though its community guidelines banned nudity. (“While we recognize that this photo is iconic, it’s difficult to create a distinction between allowing a photograph of a nude child in one instance and not others,” a spokesman for Facebook said in response to queries from the Guardian.) In the wake of the censorship of the photograph, Espen Egil Hansen, the editor-in-chief and CEO of Norway’s largest paper, would declared that Zuckerberg was “the world’s most powerful editor.

Democracy by Likes

What if Facebook put governance decisions to a vote, asking people to like or dislike a particular post? We do not typically solve controversies over a particular statement through popular vote. This may be because such a mechanism might often degenerate into a contest focused on the popularity of the controversial content, rather than a reasoned assessment of whether the content violated the community guidelines. This would have the effect of reinforcing popular views at the cost of minority viewpoints.

Feudal Lords

What of Reddit-style moderators, granted the authority to regulate particular discussions, charged with the authority to remove posts they found to be a violation of that group’s guidelines? This essentially becomes a kind of dispersed version of Mark Decides—instead of a single king, multiple lords. If only a few “lords” have power, this approach raises the same concentrated power issues of Mark Decides. If there are many “lords,” however, there might be enough alternatives that controversial content might find some home somewhere. Thus, such an approach might only shift the material to different corners of Facebook, rather than removing material that genuinely violates Facebook’s community guidelines.

Judge Robot

Perhaps we could rely upon a computer to make content decisions. In fact, of course, if Facebook is making millions of content decisions each day, it is likely relying in significant part on AI. Natasha Duarte, Emma Llansó, and Anna Loup have argued, however, that “large-scale automated filtering or social media surveillance […] results in overbroad censorship, chilling of speech and association, and disparate impacts for minority communities and non-English speakers.” As Duarte et al. describe, decision-making by AI would not result in unbiased decisions, but rather would potentially amplify bias. I have described this in my scholarship as a kind of “viral discrimination.”

Official Censorship

There is reason to worry about the great private power concentrated in companies like Facebook, which reserves the right to delete information that violate its community guidelines. An alternative is to vest such decision-making in traditional governments—either through courts or administrative bodies. Controversies over content would then be determined by an official process, rather than by corporate employees following often secret processes and secret rules. Of course, it is unclear whether individuals would have the resources to bring or defend claims that some material should be removed. Not only may the process be expensive, it may also be slow. Furthermore, governments may use their content management powers to target negative or opposition information as “fake news.” This concern will be heightened in illiberal states. Finally, states with a strong commitment to free expression, like the United States, would find it difficult to censor content that was not illegal except under time, place, and manner restrictions that are difficult to translate to the internet.

Afterword: Facebook’s Last Experiment with Outside Decision-Making

This is not Facebook’s first experiment with ceding decision-making to outsiders. In 2009, Facebook permitted its users to vote on proposed changes to the terms of use—though on terms that made practically all such votes advisory rather than mandatory. That Facebook democracy proved short-lived.
Under the now-defunct system, a site-wide vote would be triggered if proposals to modify Facebook’s terms of service received comments from at least 7,000 users. Then if at least 30 percent of active users voted on the proposal, Facebook would treat the vote as binding. Otherwise the vote would be merely advisory. Given that Facebook already had a billion users, there was little likelihood that any vote would be binding.

In December 2012, Facebook put two policy changes to a vote: whether Facebook should be able to share data with Instagram and whether Facebook should end users’ rights to vote on further governance questions. 88 percent of the 668,872 voters resoundingly rejected the changes.  But the vote fell far short of the 300 million or so required for a binding vote, and thus could be safely ignored.
After merely three votes, Facebook ended its chimerical experiment with democracy. Perhaps Facebook’s new experiment with external governance might prove longer-lived.

Anupam Chander is Professor of Law, Georgetown University; A.B. Harvard, J.D. Yale. The author is grateful to Delia Brennan and Ryan Whittington for superb research assistance, and also thankful for a Google Research Award to support related research. You can reach him by e-mail at ac1931 at georgetown.edu