Oversight Board Overturns Meta's Original Decision in Holocaust Denial Case
The Oversight Board has overturned Meta’s original decision to leave up an Instagram post containing false and distorted claims about the Holocaust. The Board finds that the content violated Meta’s Hate Speech Community Standard, which bans Holocaust denial. This prohibition is consistent with Meta’s human-rights responsibilities. The Board is concerned about Meta’s failure to remove this content and has questions about the effectiveness of the company’s enforcement. The Board recommends Meta take steps to ensure it is systematically measuring the accuracy of its enforcement of Holocaust denial content, at a more granular level.
About the Case
On September 8, 2020, an Instagram user posted a meme of Squidward – a cartoon character from the television series SpongeBob SquarePants. This includes a speech bubble entitled “Fun Facts About The Holocaust,” which contains false and distorted claims about the Holocaust. The claims, in English, question the number of victims of the Holocaust, suggesting it is not possible that six million Jewish people could have been murdered based on supposed population numbers the user quotes for before and after the Second World War. The post also questions the existence of crematoria at Auschwitz by claiming the chimneys were built after the war, and that world leaders at the time did not acknowledge the Holocaust in their memoirs.
On October 12, 2020, several weeks after the content was posted, Meta revised its Hate Speech Community Standard to explicitly prohibit Holocaust denial or distortion.
Since the content was posted in September 2020, users reported it six times for violating Meta’s Hate Speech policy. Four of these reports were reviewed by Meta’s automated systems that either assessed the content as non-violating or automatically closed the reports due to the company’s COVID-19 automation policies. These policies, introduced at the beginning of the pandemic in 2020, automatically closed certain review jobs to reduce the volume of reports being sent to human reviewers, while keeping open potentially “high-risk” reports.
Two of the six reports from users led to human reviewers assessing the content as non-violating. A user who reported the post in May 2023, after Meta announced it would no longer allow Holocaust denial, appealed the company’s decision to leave the content up. However, this was also automatically closed due to Meta’s COVID-19 automation policies, which were still in force in May 2023. They then appealed to the Oversight Board.
The Board finds that this content violates Meta’s Hate Speech Community Standard, which prohibits Holocaust denial on Facebook and Instagram. Experts consulted by the Board confirmed that all the post’s claims about the Holocaust were either blatantly untrue or misrepresented historical facts. The Board finds that Meta’s policy banning Holocaust denial is consistent with its human-rights responsibilities. Additionally, the Board is concerned that Meta did not remove this content even after the company changed its policies to explicitly prohibit Holocaust denial, despite human and automated reviews.
As part of this decision, the Board commissioned an assessment of Holocaust denial content on Meta’s platforms, which revealed use of the Squidward meme format to spread various types of antisemitic narratives. While the assessment showed a marked decline since October 2020 in content using terms like “Holohoax,” it found that there are gaps in Meta’s removal of Holocaust denial content. The assessment showed that content denying the Holocaust can still be found on Meta’s platforms, potentially because some users try to evade enforcement in alternative ways, such as by replacing vowels in words with symbols, or creating implicit narratives about Holocaust denial using memes and cartoons.
It is important to understand Holocaust denial as an element of antisemitism, which is discriminatory in its consequences.
The Board has questions about the effectiveness and accuracy of Meta’s moderation systems in removing Holocaust denial content from its platforms. Meta’s human reviewers are not provided the opportunity to label enforcement data in a granular way (i.e., violating content is labelled as “hate speech” rather than “Holocaust denial”). Based on insight gained from questions posed to Meta in this and previous cases, the Board understands these challenges are technically surmountable, if resource intensive. Meta should build systems to label enforcement data at a more granular level, especially in view of the real-world consequences of Holocaust denial. This would potentially improve its accuracy in moderating content that denies the Holocaust by providing better training materials for classifiers and human reviewers. As Meta increases its reliance on artificial intelligence to moderate content, the Board is interested in how the development of such systems can be shaped to prioritize more accurate enforcement of hate speech at a granular policy level.
The Board is also concerned that, as of May 2023, Meta was still applying its COVID-19 automation policies. In response to questions from the Board, Meta revealed that it automatically closed the user’s appeal against its decision to leave this content on Instagram in May 2023, more than three years after the pandemic began and shortly after both the World Health Organization and United States declared that COVID-19 was no longer a “public health emergency of international concern.” There was a pressing need for Meta to prioritize the removal of hate speech and it is concerning that measures introduced as a pandemic contingency could endure long after circumstances reasonably justified them.
The Oversight Board’s Decision
The Oversight Board overturns Meta’s original decision to leave up the content.
The Board recommends that Meta:
- Take technical steps to ensure that it is sufficiently and systematically measuring the accuracy of its enforcement of Holocaust denial content, to include gathering more granular details.
- Publicly confirm whether it has fully ended all COVID-19 automation policies put in place during the pandemic.
For Further Information
To read the full decision, click here.
To read a synopsis of public comments for this case, please click the attachment below.