<![CDATA[The present study reports the design, development, and evaluation of DeepCrit, a deep learning (DL)–driven intelligent tutoring system (ITS) intended to enhance students’ critical action skills in science education. While ITS research has expanded rapidly alongside advances in adaptive learning and learner modeling, few systems explicitly target higher-order skills such as critical action—a dimension of critical consciousness involving the capability to analyze socio-scientific issues, design evidence-based solutions, and enact transformative actions. This study addresses this gap through a multiphase mixed-method evaluation integrated with Design-Based Research (DBR). The research involved preliminary needs analysis, conceptual design, prototype development, expert validation, and classroom implementation with 62 secondary school students. DeepCrit integrates deep knowledge tracing, multi-task learner profiling, a knowledge graph–based domain model, and a pedagogical engine driven by a deep Q-network to provide adaptive and dialogic scaffolding around socio-scientific issues. Quantitative results demonstrate significant improvements in critical reflection, critical motivation, and critical action, alongside gains in conceptual science mastery. Qualitative findings reveal that DeepCrit supports students’ movement through praxis cycles—reflection, decision-making, and action—thereby strengthening their scientific agency. This study contributes a pedagogical and technical framework for designing ITS that support transformative science learning aligned with the demands of the 21st century.]]>
Copyrights © 2026
