Design Principles for Enhancing AI-Assisted Moderation in Hate Speech Detection on Social Media Platforms

Hate speech on social media poses a growing threat to individuals and society, necessitating technological support for moderators in detecting and addressing problematic content. This article explores the design principles essential for creating effective user interfaces (UIs) in decision support systems that employ artificial intelligence (AI) to aid human moderators. Through a comprehensive study involving 641 participants across three design cycles, we qualitatively and quantitatively evaluate various design options. Our assessment encompasses perceived ease of use, usefulness, and intention to use, while also delving into the impact of AI explainability on users' cognitive efforts, informativeness perception, mental models, and trustworthiness. Notably, software developers affirm the high reusability of the proposed design principles. The findings reveal that well-designed UIs can significantly enhance the effectiveness of AI-based moderation tools, providing clear and understandable explanations that improve user trust and engagement. By addressing both technical and user-centered aspects, this research contributes to the development of more robust and user-friendly AI systems for hate speech detection. Future work should focus on further refining these principles and exploring their applicability in diverse social media contexts to ensure comprehensive and adaptable solutions for content moderation.