The Facebook Oversight Board: Some thoughts on transparency
Since it published its first six decisions this January, the Oversight Board established by Facebook has begun to attract (modest) scholarly attention, boosted by the publication of its decision in the case of Donald Trump’s ‘deplatforming’ last week (Decision 2021-001-FB-FBR of May 5, 2021). This court-like body uses the rhetoric of constitutional justice, but it is indeed a private organ, which raises several interesting questions. Can we draw an analogy between Facebook’s Community Standards (coupled with the Oversight Board’s foundational governing document, the Charter) and a constitution? Can we draw an analogy between the Oversight Board and a constitutional court specialized on freedom of expression? And if we can, how far these analogies may go? It had been envisioned as ‘almost like a Supreme Court’ by Mark Zuckerberg himself back in 2018 and consequently often referred to as the ‘Facebook Supreme Court’ in the media. Scholars have described it as a quasi-court. The Board, instead, defines itself as a ‘non-judicial grievance mechanism’ (for the first time in Decision 2020-006-FB-FBR, Section 8.3, and then again in the Trump case, see Sec. 3).
The Board’s motto, which appears in the title of its website and in the header of its decisions, is ‘Independent judgment. Transparency. Legitimacy’. The Charter solemnly proclaims that ‘the board will operate transparently, and its reasoning will be explained clearly to the public’. At the same time, the Bylaws (the Board’s secondary governing document that fleshes out the provisions of the Charter) provides that deliberation is to ‘be held privately to protect the information the board is reviewing and the security of the panel members’ (Art. 1, § 3.1.6.). Nothing surprising, if we accept the analogy with courts. Less conventional, instead, is the provision that makes the composition of the adjudicating five-member panels confidential (Charter, Art. 3, Sec. 2). Although the membership of the entire Oversight Board is public, the single members (never called ‘judges’ in any of the foundational documents) remain anonymous when deciding cases. The Charter actually specifies the reason behind this choice in the very same provision: ‘to ensure the safety and independent judgment of panel members’ (repeated in the Bylaws, Art. 1, § 3.1.3.). This might be an understandable choice from a civil law perspective, as in civil law countries we are used to judges who remain in the shadows and judicial decisions are delivered in the name of the court (or, often, of the state). It is, however, a more questionable choice from a common law perspective, which is the perspective of lawyers educated in the United States, the home country of Facebook. All the more interesting is that the US-based Stanford Law and Policy Lab recommended exactly this solution ‘to protect board member deliberation’, but without elaborating further on this matter (see pp. 5 and 36 of their Recommendations for the Facebook Content Review Board issued in 2019). One may wonder whether the choice of anonymous decisions confirms the non-judicial nature of the Oversight Board or rather represents a preference for a civil law approach to adjudication.
The secret composition of the panels may be seen as problematic for transparency, even though it is mitigated by the fact that four of the five members are assigned randomly by an automated system (Bylaws, Art. 1, § 3.1.3.). The fifth member shall come from the region where the case arose (Charter, Art. 3, Sec. 2), but even this member is assigned randomly if there are more members from the given region. In relation to this it can be mentioned as, to put it naively, a curiosity that notwithstanding the declared intent of ensuring diversity of composition (Charter, Art. 1), one-fourth of the members currently sitting on the Oversight Board (five of 20) come from the United States. Even assuming that the automated system of assignment ensures regional diversity on the panels, it is mathematically very likely that every panel will include a member that comes from the United States. However, we cannot know this for sure, given the confidentiality of the panels’ composition, and this compromises the credibility of the Board and the seriousness of its strive for diversity.
Still related to the transparency of decision-making, another unique solution adopted by Facebook was to allow the Board to disclose dissent but without breaking the adjudicating panel’s anonymity. From a comparative perspective, anonymous judicial dissent is not completely new but it is definitely exceptional. The Greek solution deserves special mention. Interestingly, anonymity of dissent in Greek courts is coupled with another unique rule, which makes disclosure of dissent compulsory (a rule enshrined in Art. 93(3) of the Greek Constitution). No similar imposition exists, however, for the members of the Oversight Board. While the rationale behind the anonymity rule is to avoid putting the individual members in the spotlight, compulsory disclosure of dissent certainly makes this reason more pressing.
As the Oversight Board has only published ten decisions so far, we cannot talk about consolidated practice, but these ten cases allow us to make some observations and to identify questions to think further. Just as it is done in Greece, the Oversight Board has made minority opinions an integral part of the reasoning instead of publishing them separately. This means that the minority opinion is (presumably) not authored by the minority itself, but it is phrased by the opinion-writer and becomes part of a dialectical reasoning. This practice is not in conformity with the Stanford Lab’s recommendations, which called for the possibility of ‘a dissenting or concurring opinion to accompany the final decision’ (see p. 32). It is, however, consistent with the Charter’s dedication to consensus-based decision-making (see Art. 3, Sec. 4). At the same time, four of the ten decisions present minority views. Considering that the panels are composed of five members only, 40% is a high rate. It is also remarkable that among the first six decisions that were published altogether on January 28 only one was non-unanimous. Was it a careful choice of the Oversight Board to publish mostly unanimous decisions at its première? It would not be surprising, knowing the first decades of history of the US Supreme Court and Chief Justice Marshall’s struggle for unanimous decisions. In any case, of the four cases decided later, three present minority views. A very different proportion.
Since the minority’s views are not formulated as a separate opinion but incorporated in a dialectical reasoning, they are not labelled as dissent or concurrence. However, if we look into the content of those minority views, we may see that in the first three of the four non-unanimous cases they may be characterized as dissenting opinions. In these cases, the minority disagreed with the outcome and would have either overturned Facebook’s decision (in the Azerbaijani case) or would have upheld it (in the Indian Muslims’ case and the Dutch ‘Zwarte Piet’ case), to the contrary of the majority. It is never specified whether the minority consisted of one or two members of the five-member panel, but in the Azerbaijani case the reasoning suggests that there might have been two dissenters. In any case, the panels’ decisions are not final but have to be approved by a majority of the entire Board (Charter, Art. 3, Sec. 7.1). Thus, it is all the more interesting to see that non-unanimous opinions also get the approval, even though there is no public information about whether any of the ten cases has been sent back to a new panel for re-review (Bylaws, Art. 1, § 3.1.8.).
The Board’s decision in Trump’s case and the minority opinions there presented are particular in several ways. It is the first case where the minority opinion may be defined as a concurring opinion. Moreover, the reasoning is remarkably longer than in the previous cases, also due to the presence of four different minority views. These four different views may have been expressed by four different members or they may all come from the same member. We do not know, and there is nothing in the reasoning that would offer a clue about the number of members behind these minority views. The first two concurring views claim that there are additional grounds for the decision, the third one supplements the majority’s proportionality analysis, while the fourth one is related to remedial action. Thus, a minority argument is expressed in relation to four distinct questions. This suggests that there must have been a very lively debate in the panel before reaching a decision.
At last but not least, a substantive conclusion that we may draw from reading the minority opinions is that the proportionality test seems to be the most central and debated issue for the Oversight Board. In all four non-unanimous cases at least part of the minority view concerned the proportionality analysis.
The Oversight Board’s ambitious motto ‘Independent judgment. Transparency. Legitimacy’ poses a great challenge to this newly established body. There is a clear tension between the first two of these values. Indeed, it seems to be the underlying reason for the confidentiality of the panels’ composition that publicity may endanger the independence of the single members, and consequently of the Board as a whole. As regards legitimacy, instead, a recent empirical study found that in cases of higher political salience the formulation of dissenting opinions can be a meaningful way of securing greater support for a court’s policy outputs by suggesting evidence of procedural justice (Bentsen 2019). Therefore, it seems to be a wise solution to allow the disclosure of minority opinions, enhancing the Oversight Board’s legitimacy and public acceptance, but in a way that does not jeopardize the independence of the single members. This is an especially delicate question in relation to a newly established adjudicating body, which still has to build up its reputation.
On a final note, notwithstanding the questionable transparency of its own decision-making process, the Oversight Board contributes to enhance the transparency of Facebook (on this, listen to the Lawfare podcast on the Trump case, in which the participants conclude that the Oversight Board’s decisions fulfill an important ‘information-forcing function’). In its decisions, the Board repeatedly called for a greater transparency of how Facebook makes its content removal decisions. Matthew Schafer, a media law scholar, argued that the Board falls short of the very same principle that it applies to Facebook: access to information. He bases his criticism on the confidentiality of the panels’ composition and on the fact that the Oversight Board does not publish briefing provided by Facebook. But while the latter could be remedied by the Board, confidentiality of the panels’ composition has been imposed on the Board by Facebook through the Charter. Moreover, a balance between independence and transparency is difficult to draw. At the end of day, the question is whether revealing the identity of the members of the panel and of the dissenters is worth the potential negative effect that this could have on the Board’s independence.