<![CDATA[This study aims to develop and validate a deep learning-based framework for measuring affective outcomes in Islamic Religious Education (IRE) through the integration of multimodal learning data, qualitative validation, and ethical governance grounded in Islamic values. Employing a mixed-methods sequential explanatory design, the research was conducted at SMAN 24 Kabupaten Tangerang, SMAN 19 Kabupaten Tangerang, and SMA Islamic Village. The quantitative phase involved the development of Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), and transformer-based models to analyze students’ reflective texts, behavioral logs, and digital interaction patterns. The qualitative phase consisted of teacher and student interviews, classroom observations, and document analysis to contextualize and validate computational findings. The results demonstrate that deep learning models effectively measure religiosity, empathy, moral awareness, ethical disposition, and spiritual engagement in a more objective, adaptive, and sustainable manner than conventional affective assessment approaches. Multimodal models exhibited the highest predictive performance by integrating cognitive, behavioral, and affective indicators into a unified analytical framework. The study further highlights the necessity of embedding principles of adl (justice), amanah (trustworthiness), data privacy, and informed consent into AI-based educational assessment systems. Overall, the proposed framework contributes to the advancement of human-centered, ethically grounded, and data-driven affective assessment in the digital transformation of Islamic education.]]>
Copyrights © 2026
