Operationalizing the legal concept of 'Incitement to Hatred' as an NLP task

Hate speech detection or offensive language detection are well-established but controversial NLP tasks. With 'hate speech' not being a legal term, these tasks often elaborate on the question of which statements are perceived as being 'appropriate' or 'offensive'. Looking beyond this cursory understanding, this article proposes a way to operationalize the legal concept of incitement to hatred as an NLP task. We pursue this task based on the criminal offense of incitement to hatred in {\S} 130 of the German Criminal Code along with the underlying EU Framework. Under the German Network Enforcement Act, social media providers are subject to a direct obligation to delete postings violating this offense. We take this as a use case to study the transition from the ill-defined concepts of hate speech or offensive language which are usually used in NLP to an operationalization of an actual legally binding obligation. We first translate the legal assessment into a series of binary decisions and then collect, annotate, and analyze a dataset according to our annotation scheme. Finally, we translate each of the legal decisions into an NLP task based on the annotated data. The two subtasks of target group detection and targeting act detection can be considered crucial from a legal viewpoint. We show that both can be annotated by non-legally trained persons with sufficient reliability.