An independent oversight board that reviews content moderation decisions at Meta has suggested that the company revise its cross-check program, and the company has agreed — sort of.
In total, The Oversight Board, the “independent body” that reviews Meta’s content moderation decisions, issued 32 suggestions for amending the program, which places content from “high-profile” users in a moderation queue separate from the automated one the company uses for normies. Instead of being taken down, flagged content from select public figures like politicians, celebrities, and athletes is left up “pending further human review.”
The Board’s review was conducted in direct response to a 2021 Wall Street Journal article that examined the exempted. In their decision, the board acknowledged the inherent challenges of moderating content at scale, saying that though “a content review system should treat all users fairly,” the program grapples with “broader challenges in moderating immense volumes of content.”
For example, at the time of the request, they say Meta was performing such a high volume of daily moderation attempts — about 100 million — that even “99% accuracy would result in one million mistakes per day.
Still, the Board says the cross-check program was less concerned with “advanc[ing] Meta’s human rights commitments” and “more directly structured to satisfy business concerns.”
Of the 32 suggestions the Board proposed to amend the cross-check program, Meta agreed to implement 11, partially implement 15, continue to assess the feasibility of one, and take no further on the remaining five. In an updated blog post published Friday, the company said it would make the program “more transparent through regular reporting,” as well as fine-tune criteria for participation in the program to “better account for human rights interests and equity.” The company will also update operational systems to reduce the backlog of review requests, which means harmful content will be reviewed and taken down more quickly.
All 32 recommendations can be accessed at this link.
The Board noted in its Twitter thread that the changes “could render Meta’s approach to mistake prevention more fair, credible and legitimate” but that “several aspects of Meta’s response haven’t gone as far as we recommended to achieve a more transparent and equitable system.”