Deradicalizing YouTube: Characterization, Detection, and Personalization of Religiously Intolerant Arabic Videos

Growing evidence suggests that YouTube's recommendation algorithm plays a role in online radicalization via surfacing extreme content. Radical Islamist groups, particular, have been profiting from the global appeal of YouTube to disseminate hate and jihadist propaganda. In this quantitative, data-driven study, we investigate prevalence religiously intolerant Arabic videos, tendency platform recommend such how these recommendations are affected by demographics watch history. Based on our deep learning classifier developed detect hateful videos large-scale dataset over 350K find targeting religious minorities particularly prevalent search results (30%) first-level (21%), 15% overall captured point videos. Our personalized audit experiments suggest gender identity can substantially affect extent exposure contribute vital insights into phenomenon facilitate curbing harmful