<![CDATA[The growing demand for 21st-century competencies has driven major transformations in basic education; however, a clear research gap persists regarding the conceptual ambiguity between deep learning as a pedagogical approach and deep learning as artificial intelligence (AI) technology, as well as the unequal implementation of both across educational contexts. This study aims to systematically analyze and compare how deep learning is conceptualized and applied in basic education globally through a Systematic Literature Review (SLR) guided by the PRISMA protocol. From 50 peer-reviewed articles published between 2020 and 2025, 13 high-quality studies were selected based on relevance, methodological rigor, and contextual focus. The findings reveal a distinct contextual divergence. In Indonesia, deep learning is predominantly implemented as a pedagogical strategy emphasizing mindful, meaningful, and joyful learning to support character development within the Merdeka Curriculum. In contrast, studies from developed countries primarily frame deep learning as an AI-based technological application, particularly for learning performance prediction and inclusive education support using neural network models such as Convolutional Neural Networks (CNN) and Bidirectional Long Short-Term Memory (BiLSTM). While both approaches demonstrate effectiveness in fostering higher-order thinking skills (HOTS), their successful implementation in developing countries is strongly constrained by disparities in digital infrastructure and teacher competency readiness. This study contributes by clarifying the dual interpretation of deep learning in basic education and highlighting the contextual factors shaping its implementation. Accordingly, it recommends the development of adaptive teacher training models and more equitable technology integration policies to bridge global disparities in educational quality.]]>
Copyrights © 2026
