Monday, July 18, 2022

Meta’s Oversight Board Often Turns in Its Homework Late. Does it Matter?

By evelyn douek, Tia Sewell Friday, July 15, 2022


A sign outside of Meta headquarters in California. 

On June 17, the Meta Oversight Board released its 25th decision. This decision, in the Knin Cartoon case, overturned Meta’s original decision to leave up an edited cartoon video that depicted ethnic Serbs as rats. According to the operating rules that Meta wrote when setting up the board, this decision was due for release to the public on June 13. That is, the board’s decision was four days overdue, according to its bylaws. This may seem like a pedantic nitpick. Four days? It seems unlikely anyone lost sleep over this small delay.

But the delay is worth paying attention to, precisely because it says something interesting about a deeper question that surrounds the board: What kind of institution is it, exactly?

This post sets out the board’s somewhat frequent record of releasing decisions late, in breach of its own constituent documents, before examining what it might suggest about the board more broadly.

A Record of Late Decisions

The board’s bylaws read:
Beginning when the selection of a case is published, the board has a timeframe of ninety (90) days to render a final decision with respect to any particular case. This period may be extended in exceptional circumstances or in the event of technical or operational incidents which impede the Board from publishing a final decision within this target timeframe. If the timeframe is extended, that extension will be notified to the posting person (as well as the reporting person, if different), Facebook and the public. The administration, on behalf of the board, will monitor each chosen case and ensure the board issues its decision within this timeframe. (Emphasis added.)

In the Knin Cartoon case, the Oversight Board announced that it would consider the case on March 15, along with two other cases, the Sudan Graphic Video case and the Reclaiming Arabic Words case. The board issued decisions for the latter two on the date they were due—that is, June 13, exactly 90 days after they were announced. But while handing down these rulings, the board made no mention of the Knin Cartoon case. Indeed, neither then nor later did the board issue a notice of an extension to the public.

And this was not the first time, by our count, that the Oversight Board missed a deadline: Back in February, the board published two decisions—its Advice on Pharmaceutical Drugs case and its Journalism on Sexual Violence case—a day late, according to the board’s bylaws. (The decisions were delivered on Feb. 1, 91 days past their announcement date of Nov. 2, 2021.)

When Lawfare asked the Oversight Board to comment on these delays, it responded:

The Board has, in exceptional instances, delayed the issuance of certain case decisions as permitted by the organization’s bylaws. Thoughtful processes and the incorporation of diverse perspectives are critical to effective recommendations and the Board’s ability to drive greater transparency and accountability at Meta. While the Board seeks to balance the swift delivery of case recommendations, it will not sacrifice rigor for the sake of expediency.

The board is right to note that the bylaws allow case decisions to be delayed in the event of “exceptional circumstances.” And we definitely don’t want to suggest we are asking it to “sacrifice rigor”! But the bylaws explicitly say that in such an event, the board should notify the public. The board is clearly aware of this requirement and sees its value. It has, in the past, provided such notification: For example, the board took 104 days to render its final decision in the Trump Suspension case—a delay in conformity with its bylaws, as the board invoked the “exceptional circumstances” clause and provided public notification on April 16, 2021, shortly before the original case deadline.

In other cases, the board has provided post hoc notification of delayed decisions. According to the board’s most recent transparency report, published in June, three additional decisions were overdue: The Wampum Belt and Ayahuasca Brew rulings were handed down 99 days after the beginning of the 90-day period (nine days late) and the Alleged Crimes in Raya Kobo ruling came 104 days after the period began (14 days late). In this report, footnotes explained that “[f]or this quarter, all three case decisions experienced delays resulting in decision publication and implementation going beyond the regular 90-day period. These delays were caused by availability issues both within Meta and the Oversight Board due to season holidays.” Further, the board added that “[t]here was an additional delay in the Ethiopia case due to translation issues.”

Lawfare’s online tracker had not caught these delays, as the board began its countdown when the cases were assigned to panel—a date that was not publicly available until the transparency report was released—rather than when the case was announced, which is the date that Lawfare has used to track deadlines. Surely such post hoc notification of delayed decisions violates the spirit, if not the letter, of the board’s rules.

In March 2021, the board had revised its bylaws to require that the 90-day timeframe begin “when the Board assigns a case to panel” as opposed to the date of “Facebook’s last decision on a case.” In November 2021, the board again changed its bylaws to initiate the 90-day clock “when we publish our selection of a case, rather than when it is assigned to panel.”

All of this is to say, six of the 25 case decisions the board has released so far have been late without comment provided prior to the deadline, as shown in Figure 1. That is, more than 20 percent of the board’s decisions are late, in breach of the board’s own bylaws.

A Flourish chart
Figure 1. Overview of Oversight Board case timelines, with cases ordered sequentially by decision date. Data sourced from Oversight Transparency reports through Q4 2021; Lawfare tracker for decisions thereafter. (*) Notification of delay provided to public in advance of decision.

Why Does This Matter?

The most obvious thing that this record of late decisions suggests is that the board is somewhat overwhelmed by its workload. Decisions are issued consistently right up against their deadline and—as Figure 1 suggests—this is happening more and more over time. If the board is struggling to keep up with its current caseload—a modest six cases have been announced in the past eight months—this suggests that the board’s current docket size is about its limit. With the board only ever being able to address a minuscule fraction of the content moderation issues that arise across Meta’s platforms, this could be a massive constraint on the board’s capacity to make an impact. (For context, more than 1.1 million cases were submitted to the board by users and Meta from October 2020 to December 2021. Also perhaps relevant context: Board members are reportedly earning six-figure salaries.)

This makes it all the more puzzling that the board has not increased its size in a meaningful way. According to the board’s charter and bylaws, it has an “ideal size” of 40 members when “fully staffed.” The board announced its first 20 members over two years ago, added (and lost) one member in 2021, and has most recently increased its size by announcing another three members in May of this year. The board has not explained why it has taken so long to ramp up its staffing (although presumably the pandemic played a role). But this is clearly something the board could have, and should have, more actively prioritized to maximize its potential—given what the delays suggest about the role of bandwidth in limiting the board’s output.

The board’s late decisions also highlight its weird institutional status. The board is a purely self-regulatory, private body, set up by then-Facebook, with no independent legal authority or particular source of constitutive legitimacy. Indeed, for exactly this reason many observers pronounced the board a distracting charade. Seemingly in order to counteract these perceptions, Facebook and the board dressed up the board in all the trappings of a court-like institution. It was established by a charter, and it has bylaws. Its decisions clearly imitate judicial opinions—citing legal authorities and precedents, and noting dissenting opinions among its members. The attempted message of all these features is clear: The board is a serious institution that you should take seriously. Just like judges wearing robes, the board is dressed up in official-looking clothing to attempt to manifest authority.

But if this legitimacy is to be sustained, the board needs to commit to the bit even when it’s inconvenient. And failing to comply with the undemanding requirement of providing public notice of late decisions does not inspire confidence about the board’s commitment to its other operating principles, many of which cannot be verified by external observers.

Again, we don’t want to overstate the importance of a couple of missed deadlines. Goodness knows we’ve missed a bunch of deadlines of our own. But if the board wants others to take its performance seriously, maybe it should lead by example and take its rules seriously itself.

evelyn douek
evelyn douek is an Assistant Professor of Law at Stanford Law School and Senior Research Fellow at the Knight First Amendment Institute at Columbia University. She holds a doctorate from Harvard Law School on the topic of private and public regulation of online speech. Prior to attending HLS, evelyn was an Associate (clerk) to the Honourable Chief Justice Susan Kiefel of the High Court of Australia. She received her LL.B. from UNSW Sydney, where she was Executive Editor of the UNSW Law Journal.
Tia Sewell
Tia Sewell is an associate editor of Lawfare. She is an undergraduate at Stanford University studying international relations and economics.

No comments:

Post a Comment