I recently published a working paper that examines the effectiveness of Meta’s Oversight Board in driving pro-expression norms and principles by analyzing over 100 Board cases. Overall, I find that the Board has had mixed results in advancing free expression on Meta’s platforms.
This outcome was not what I would have expected if you had asked me five years ago. Back then, I was working on Meta’s content policy team, and I signed up to support the newly established Board because I felt that creating a diverse team of thinkers to protect free expression could only improve the content moderation decisions being made not only by Meta, but even the broader Trust and Safety community. I hoped that the Board, free from the Silicon Valley monoculture, immune to the considerations of advertising boycotts, and with a better—though certainly not perfect—balance of perspectives, would challenge prevailing norms that restricted important political or social speech, leading to more open expression online.
However, my research reveals that the Board’s impact is that it protected expression in some places but restricted it in others. Specifically, I found that the Board advanced expression through its ability to find errors in Meta’s content moderation. The Board has rectified errors in dozens of cases, mostly to restore speech. But while this has improved expression, it’s still fixing a tiny fraction of the total number of errors referred to the Board and an even smaller fraction of the total number of errors made by Meta.
The Board has also improved expression through its efforts to improve clarity and transparency in Meta’s policies. Improvements to transparency can help users better understand and follow Meta’s rules and avoid having content removed. But not all transparency recommendations lead to meaningful changes.
But the Board’s biggest opportunity to expand expression—creating broad norms for free expression in content policies—has been far more mixed. In those cases that were focused on setting policy norms, about 60 percent expanded expression and 40 percent restricted expression. But those decisions to restrict expression were generally made more forcefully and set more far-reaching precedents.
Furthermore, the Board’s decisions have not protected expression equally. When the Board took cases with left-coded speech, such content was allowed to remain online 90 percent of the time, while right-coded speech was allowed 41 percent of the time. The inconsistency in protecting speech, especially for certain viewpoints, is a problem for an organization conceived as a protector of expression online.
In sum, my working paper shows that while the Board has had some success in fixing mistakes and improving transparency, the Board has not successfully been a consistent advocate for expression in its decisions and norms.
I offer several ideas for how the Board might reform itself, including a greater emphasis on free expression in future Board members and the Board staff, making it harder to pass speech-restrictive recommendations, or even pushing for greater user controls. Of course, it is up to the Board and Meta to decide what policies govern their platforms, free from government coercion or regulation. But we would all benefit from the Board doing more to strengthen a culture of free expression online.
