Can NLP Tackle Hate Speech in the Real World? Stakeholder-Informed Feedback and Survey on Counterspeech (arXiv)

Responding to hate speech online, or counterspeech, has become a popular and promising NLP intervention. Recent research frequently uses automated techniques and legacy datasets with little participation from impacted communities, whereas early studies collaborated with NGOs. The current study examines how stakeholder involvement influences dataset creation, model development, and evaluation by reviewing 74 NLP works on counterspeech. Effective stakeholder-informed practices are highlighted in a participatory case study involving five NGOs that focuses on online gender-based violence (oGBV). Recommendations for re-centering stakeholder expertise in counterspeech activities stem from findings that indicate a widening gap between NLP research and the requirements of affected communities.

https://arxiv.org/abs/2508.04638