UN Special Rapporteur on Minority Issues
[DRAFT FOR GLOBAL CONSULTATION]
such as which moderation response was taken beyond removal such as warning labels, limiting
engagement, sharing, geo-blocking, account limitations.
10. Independent, academic and civil society actors should be allowed meaningful
access to data to collaborate with SMCs on verifying and gauging effectiveness of
content moderation of hate speech particularly against minorities with a focus on
how much hateful content is not removed and why, as well as community
members’ experiences and perceptions.
Commentary
The current models for gauging effectiveness of hate speech content moderation focus on content
removals. These do not cater for hateful content that is not detected and receives no complaints. They
also do not factor in wrongful removal by automated systems that were not appealed. Both of these
occurrences are worsened if community members are not aware of the hate speech policy, or do not
know how it is applied and do not know how to flag such content.
In working with third party independent researchers, SMCs can work towards independently verifying
the effectiveness of their content moderation systems by allowing for a more detailed and nuanced
picture where it may not be working. It will also lead to developing methodology and a multidimensional approach to measuring harm that goes beyond views, but also reflects virility and severity.
It will allow for more authoritative and accurate research into the linkages of online hate and offline
hate and violence and to understand how to disrupt such causation.
15