Meta should clarify the Hate Speech Community Standard and the guidance provided to reviewers, explaining that even implicit references to protected groups are prohibited by the policy when the reference would reasonably be understood. The Board will consider this recommendation implemented when Meta updates its Community Standards and Internal Implementation Standards to content reviewers to incorporate this revision.
Our commitment: We will update our Community Standards and internal policy guidance to clarify our approach to implicit hate speech, by which we mean hate speech that requires context to interpret. We will explain that it will only be removed when we can reasonably understand the user’s intent.
Considerations: We believe that people use their voice and connect most freely when they don’t feel attacked on the basis of who they are. For that reason, we don’t allow hate speech on our platforms. However, in order to review the volume of expression that people who use our platforms share every day, we have to apply a high-capacity, high-consistency approach. This is why we only remove content when we can reasonably conclude that it contains a hate speech attack based on the context.
Our reviewers make decisions based on the letter of our hate speech policy without adding their own viewpoints. Our at-scale reviewers may escalate questions for expert review where the application of our policy is ambiguous or requires additional context. Content that may contain implicit hate speech, which often uses ambiguous language or requires additional context to interpret, may qualify for this kind of escalation.
The challenges of addressing implicit hate speech can be illustrated with these examples: One post shows an image of a Neo-Nazi rally with the caption, “wow.” A second post shows a meme with two images side-by-side, one showing Adolf Hitler surrounded by German children and the other showing Angela Merkel surrounded by Syrian refugees with a caption that reads, “Germany then vs. Germany now.” In both cases, the intent behind the content is ambiguous: it could be hate speech or it could be speech that condemns hate. When we are unable to determine whether a person’s speech is hate speech, we leave that expression on our platforms.
Next steps: We will add additional language to our Community Standards and policy guidance to clarify that our approach to implicit hate speech is to remove it if it is escalated and we can reasonably understand the user’s intent. We will report on our progress in a future Quarterly Update.