12/20/23, 4:12 PM Oversight Board | Independent judgement. Transparency. Legitimacy. matching content is immediately removed. Videos added to the escalations bank after the incident included footage that showed visible human innards (which the video at issue in this case did not). Other teams at Meta were invited to refer potentially similar videos to the policy team. The policy team would then determine if they should also be added to the escalations bank. Three days after the attack, Meta added a video almost identical to the content in this case to the escalations bank. As a result, Meta's systems compared that video to content already on the platform to check for matches. While this retroactive review was taking place, the user edited their original post, adding an English-language caption to the video. It states that the church was attacked by gunmen, that multiple people were killed and describes the incident as "sad". The caption included a large number of hashtags. The majority of these were about the liveaction game "airsoft" (where teams compete to tag each other out of play using plastic projectiles shot with mock weapons). Another, according to Meta, alluded to the sound of gunfire and is also used to market firearms. Other hashtags referenced people who collect firearms and firearm paraphernalia, as well as military simulations. Shortly after the caption was added, the escalations bank's retroactive review matched the user's post to the recently added near-identical video and removed it from the platform. The user appealed. A human moderator reviewed the content and maintained the removal decision. The user then appealed to the Board. At this point, the three reports that users had made on the content had still not been reviewed and were closed. Meta told the Board that the reports had mistakenly been assigned to a lowpriority queue. In response to the Board selecting this case, Meta reviewed the near-identical video that had been placed in the escalations bank. Meta determined that it did not violate any policies because there were no "visible innards" and no sadistic caption, and removed it from the bank. However, Meta maintained its decision to remove the content in this case, as it stated that, while the narrative about the event and user's expression of sadness were not violating, the hashtags in the caption added by the user violated multiple policies. In response to questions from the Board, Meta analysed the user's posting history and found that the user had included similar hashtags on many of their recent Instagram posts. https://www.oversightboard.com/decision/IG-OZNR5J1Z/ 6/26