Explainable AI for Detecting Harassment and Extremism in Voice Chat

作者：Uzair Aslam Bhatti, Yonis Gulzar, Yazeed Yasin Ghadi, Vаlisher Sаpаyev, Odilbek Kosimov, Shaxnoza Sultanova, Khayrulla Urozboev · 年份：2026 · DOI：10.4018/979-8-2600-1664-0.ch003 · 研究领域：Hate Speech and Cyberbullying Detection、Misinformation and Its Impacts、Spam and Phishing Detection

Online voice communication platforms have become central to gaming, social networking, and virtual collaboration, but they are increasingly challenged by harassment, hate speech, and extremist content. Traditional automated moderation systems often operate as black boxes, limiting transparency and user trust. Explainable Artificial Intelligence (XAI) offers a promising solution by enabling interpretable detection of harmful speech while providing clear justifications for moderation decisions. This chapter explores the integration of XAI techniques with speech recognition, natural language processing, and deep learning models to identify harassment and extremist behavior in voice chat environments. It examines methods for improving detection accuracy, reducing algorithmic bias, enhancing accountability, and supporting human moderators through transparent decision-making. The chapter also discusses ethical considerations, privacy concerns, and future research directions for creating safer and more inclusive digital communication spaces.