Integrating Augmented Reality and Deep Learning in an Interactive STEAM-Based Digital Storybook to Enhance Elementary School Students’ Science Literacy

Abstract

Science literacy among primary school students remains a significant challenge in equipping future generations to meet the demands of the 21st century. Prior meta-analyses demonstrated that Augmented Reality (AR) markedly enhances learning outcomes and student engagement, but the STEAM method yields a moderate effect on science achievement. Nonetheless, the amalgamation of AR, STEAM, and deep learning for individualised learning in scientific education remains little examined. This study aims to develop and evaluate an Interactive Digital Storybook that utilises AR and deep learning to enhance science literacy in primary schools. The research utilised the Successive Approximation Model (SAM), which prioritises iterative design, prototyping, and refinement. The development process had three steps: (1) preparation, which included needs analysis through interviews, observations, and document studies in three East Java elementary schools; (2) iterative design and development, which included storyboarding, interface design, AR feature creation, deep learning integration, and prototype testing; and (3) implementation, which included one-on-one and small group evaluations with teachers and students. Data were gathered via expert validation (content, pedagogy, technology) and user feedback. The data were gathered through expert validation in content, pedagogy, and technology, as well as a small trial with 45 children from three public elementary schools in East Java, Indonesia. The validation results showed that the project was very likely to work (Content = 92%, Pedagogy = 90%, technology = 94%). Students reported that interactive stories, AR-based 3D simulations, and gamified challenges increased their interest and motivation. These data indicate that the produced media is suitable for additional effectiveness testing in the second year.