PERCEIVED INFORMATIVENESS OF EXPLAINABLE ARTIFICIAL INTELLIGENCE: THE CASE OF HATE SPEECH IN SOCIAL MEDIA IN QATAR

Abstract

Hate speech, in its various forms, is pervasive across most social media platforms. User-generated content can profoundly affect individuals’ well-being. To address this, platforms have integrated artificial intelligence, with explainable AI to clarify the unknown aspects of AI decision-making, recommendations, and outputs for end-users. While prior research has explored XAI goals such as trustworthiness and informativeness and adopted user-centered approaches to align them with users’ expectations, however, the application of XAI in the context of Arabic hate speech in social media remains underexplored. This study investigates how Arabic-speaking individuals perceive the informativeness of XAI representation methods in an explainable Arabic hate speech detection. Guided by a systematic literature review, and a user-centered design approach, a high-fidelity prototype of an explainable Arabic hate speech detection was developed. The prototype was evaluated through a within-subjects user-study using both subjective and objective measures, including eye-tracking. Results showed that users perceived the combination of textual and visual explanations—especially visual saliency—as more informative than other XAI methods. In contrast, more complex visualizations, such as pie charts, received more visual attention and produced longer fixation durations, but did not necessarily enhance informativeness. Additionally, social media usage duration had no significant effect on perceived informativeness. These findings suggest that perceived informativeness is driven by the clarity and structure of explanations than by visual complexity or user experience. The study highlights the importance of designing concise and visually salient XAI representations, particularly for end-users interacting with AI systems in sensitive contexts such as hate speech detection.

Date of Award	2026
Original language	American English
Awarding Institution	HBKU College of Science and Engineering