Platformization hate. Patterns and algorithmic bias of verbal violence on social media.
Patterns and algorithmic bias of verbal violence on social media
Keywords:
Verbal violence, Hate speech, Algorithms, Supervised classificationAbstract
The contribution presented is an analysis of Hate Speech of tweets during the implementation of the EU Covid Certificate policy. The research question aims to understand the phenomenon of verbal violence from two perspectives: first, from the perspective of the text. What are the most frequent topics behind the verbal violence? As hypothesized in many previous works, is the phenomenon related to the circulation of fake news or is there more to it? The second follows the users' perspective: how do users profiling verbal violence emerge? The results we have reached are, from a substantive point of view, of good interest because they show us how it is possible to see a new type of online hate. However, disagreements that we have encountered in constructing an unambiguous definition of HS for the supervised algorithm leaves many questions open. Among all, that indeed the difference between HS and freedom of expression can be very thin. In the context of large social platforms, where algorithm criteria are not always explicit, and are also platform policies - this could be a problematic issue.
