First Ask Then Act (FATA): A Community-Centered Inclusive Approach for Hate Speech Detection

In the NLP field, the significant attention paid to Hate Speech (HS) detection has highlighted how difficult it is to define HS with clear boundaries, revealing its being a context-dependent phenomenon. A recent challenge in the field of HS detection is to overcome the risks of both over-moderation and under-moderation, emphasizing the need to better understand what is perceived as hateful and what is not by communities affected by abusive language. Additionally, an interest in developing more inclusive approaches that actively involve target groups has recently emerged. This shift includes an increasing focus on underrepresented languages and communities, encouraging researchers to more actively consider ethical issues. Against this backdrop, we present a position paper with a twofold aim: firstly, we propose a review of some interdisciplinary approaches adopted so far in the field of NLP related to HS and abusive language detection; secondly, we present First Ask Then Act (FATA), a collaborative approach based on the direct involvement of individuals and target communities to collect fair and informed data. FATA proposes a multidisciplinary methodology, which integrates methods from sociolinguistics, such as surveys and focus group interviews, into the NLP data gathering workflow for HS detection.