Oversight Board publishes policy advisory opinion on Meta’s cross-check program
In October 2021, following disclosures about Meta’s cross-check program in the Wall Street Journal, the Oversight Board accepted a request from the company to review cross-check and make recommendations for how it could be improved. This policy advisory opinion is our response to this request. It analyzes cross-check in light of Meta’s human rights commitments and stated values, raising important questions around how Meta treats its most powerful users.
To read the full version of our policy advisory opinion on Meta’s cross-check program, click here.
Please note: While translations of the summary of our policy advisory opinion are already available, the full opinion is currently only available in English. Translations into other languages are underway and will be uploaded to our website as soon as possible in 2023.
As the Board began to study this policy advisory opinion, Meta shared that, at the time, it was performing about 100 million enforcement attempts on content every day. At this volume, even if Meta were able to make content decisions with 99% accuracy, it would still make one million mistakes a day. In this respect, while a content review system should treat all users fairly, the cross-check program responds to broader challenges in moderating immense volumes of content.
According to Meta, making decisions about content at this scale means that it sometimes mistakenly removes content that does not violate its policies. The cross-check program aims to address this by providing additional layers of human review for certain posts initially identified as breaking its rules. When users on Meta’s cross-check lists post such content, it is not immediately removed as it would be for most people, but is left up, pending further human review. Meta refers to this type of cross-check as “Early Response Secondary Review” (ERSR). In late 2021, Meta broadened cross-check to include certain posts flagged for further review based on the content itself, rather than the identity of the person who posted it. Meta refers to this type of cross-check as “General Secondary Review” (GSR).
In our review, we found several shortcomings in Meta’s cross-check program. While Meta told the Board that cross-check aims to advance Meta’s human rights commitments, we found that the program appears more directly structured to satisfy business concerns. The Board understands that Meta is a business, but by providing extra protection to certain users selected largely according to business interests, cross-check allows content which would otherwise be removed quickly to remain up for a longer period, potentially causing harm. We also found that Meta has failed to track data on whether cross-check results in more accurate decisions, and we expressed concern about the lack of transparency around the program.
In response, the Board made several recommendations to Meta. Any mistake-prevention system should prioritize expression which is important for human rights, including expression of public importance. As Meta moves towards improving its processes for all users, the company should take steps to mitigate the harm caused by content left up during additional review, and radically increase transparency around its systems.
The Board recognizes that the volume and complexity of content posted on Facebook and Instagram pose challenges for building systems that uphold Meta’s human rights commitments. However, in its current form, cross-check is flawed in key areas which the company must address:
Unequal treatment of users. Cross-check grants certain users greater protection than others. If a post from a user on Meta’s cross-check lists is identified as violating the company’s rules, it remains on the platform pending further review. Meta then applies its full range of policies, including exceptions and context-specific provisions, to the post, likely increasing its chances of remaining on the platform. Ordinary users, by contrast, are much less likely to have their content reach reviewers who can apply the full range of Meta’s rules. This unequal treatment is particularly concerning given the lack of transparent criteria for Meta’s cross-check lists. While there are clear criteria for including business partners and government leaders, users whose content is likely to be important from a human rights perspective, such as journalists and civil society organizations, have less clear paths to access the program.
Delayed removal of violating content. When content from users on Meta’s cross-check lists is identified as breaking Meta’s rules and while undergoing additional review, it remains fully accessible on the platform. Meta told the Board, that, on average, it can take more than five days to reach a decision on content from users on its cross-check lists. This means that, because of cross-check, content identified as breaking Meta’s rules is left up on Facebook and Instagram when it is most viral and could cause harm. As the volume of content selected for cross-check may exceed Meta’s review capacity, the program has operated with a backlog which delays decisions.
Failure to track core metrics. The metrics that Meta currently uses to measure cross-check’s effectiveness do not capture all key concerns. For example, Meta did not provide the Board with information showing it tracks whether its decisions through cross-check are more or less accurate than through its normal quality control mechanisms. Without this, it is difficult to know whether the program is meeting its core objectives of producing correct content moderation decisions, or to measure whether cross-check provides an avenue for Meta to deviate from its policies.
Lack of transparency around how cross-check works. The Board is also concerned about the limited information Meta has provided to the public and its users about cross-check. Currently, Meta does not inform users that they are on cross-check lists and does not publicly share its procedures for creating and auditing these lists. It is unclear, for example, whether entities that continuously post violating content are kept on cross-check lists based on their profile. This lack of transparency impedes the Board and the public from understanding the full consequences of the program.
The Oversight Board’s recommendations
To comply with Meta’s human rights commitments and address these problems, a program that corrects the most high-impact errors on Facebook and Instagram should be structured substantially differently. The Board has made 32 recommendations in this area, many of which are summarized below.
As Meta seeks to improve its content moderation for all users, it should prioritize expression that is important for human rights, including expression which is of special public importance. Users that are likely to produce this kind of expression should be prioritized for inclusion in lists of entities receiving additional review above Meta’s business partners. Posts from these users should be reviewed in a separate workflow, so they do not compete with Meta’s business partners for limited resources. While the number of followers can indicate public interest in a user’s expression, a user’s celebrity or follower count should not be the sole criterion for receiving additional protection. If users included due to their commercial importance frequently post violating content, they should no longer benefit from special protection.
Radically increase transparency around cross-check and how it operates. Meta should measure, audit, and publish key metrics around its cross-check program so it can tell whether the program is working effectively. The company should set out clear, public criteria for inclusion in its cross-check lists, and users who meet these criteria should be able to apply to be added to them. Some categories of entities protected by cross-check, including state actors, political candidates and business partners, should also have their accounts publicly marked. This will allow the public to hold privileged users accountable for whether protected entities are upholding their commitment to follow the rules. In addition, as around a third of content in Meta’s cross-check system could not be escalated to the Board as of May-June 2022, Meta must ensure that cross-checked content, and all other content covered by our governing documents, can be appealed to the Board.
Reduce harm caused by content left up during enhanced review. Content identified as violating during Meta’s first assessment that is high severity should be removed or hidden while further review is taking place. Such content should not be allowed to remain on the platform accruing views simply because the person who posted it is a business partner or celebrity. To ensure that decisions are taken as quickly as possible, Meta should invest the resources necessary to match its review capacity to the content it identifies as requiring additional review.
For further information
To read the Board's full policy advisory opinion, as well as Meta's request to the Board, and public comments submitted for this opinion, please click the attachments below.