Deep Learning–Based Augmented Reality for Inclusive Computational Thinking Education: A Systematic Review

Computational thinking (CT) has emerged as an essential literacy skill for 21st-century learners; however, students with disabilities remain underrepresented in STEM education. While mobile augmented reality (AR) and deep learning (DL) technologies individually hold promise for creating inclusive learning environments, their convergence in CT education remains underexplored. This systematic review examines the integration of deep learning-based mobile augmented reality systems for inclusive computational thinking education. A systematic literature review was conducted according to the PRISMA guidelines. This review revealed three significant findings. First, the integration of deep learning into mobile AR systems for CT education is in its infancy, with only one study providing empirical evidence of a fully functional DL-based augmented reality (AR) system. The vast majority of these employ static AR without adaptive capabilities. Second, a robust foundation of inclusive design principles exists, including scaffolded instruction, multimodal support, and accessible interfaces; however, these design principles have been applied predominantly to learners with cognitive and neurodevelopmental disabilities, while learners with sensory and physical impairments remain significantly underrepresented. Third, technology-mediated interventions improved CT skills and engagement across all 23 studies, but the measurement of self-efficacy was weak and predominantly qualitative, with few studies employing quantitative instruments. The three critical strands of inquiry, intelligent technology, inclusive design, and meaningful outcomes, remain disconnected. No study to date has successfully combined DL-based adaptive AR with inclusive design principles and evaluated its impact on CT skills, engagement, and self-efficacy across a broad spectrum of learners with disabilities. This review identifies gaps in the literature and proposes a roadmap for future research, emphasizing the need for interdisciplinary collaboration, broader disability representation, and the development of validated self-efficacy instruments.