Development of Deep Learning-Based Digital Interactive Multimedia to Improve Elementary Students’ Critical Thinking on Ecosystem Material

Abstract

Elementary students’ critical thinking skills are commonly reported as low under lecture-centered instruction with minimal use of interactive media, a condition also confirmed at the research site through preliminary observation showing limited use of higher-order reasoning tasks. This study aimed to develop a practical and effective deep learning-based digital interactive multimedia, RAJA MAKAN, to improve fifth-grade students’ critical thinking on ecosystem material. The product was developed using the ADDIE model (Analysis, Design, Development, Implementation, and Evaluation), a systematic and cyclical instructional design framework in which each phase produces an output that becomes the reference for the following phase, involving expert validation, formative revision, and a small-group and large-group trial with fifth-grade students. As supporting data to this development process, material and media expert validation reached 87.7% and 86.7% respectively, teacher and student practicality reached 98% and 93%, and classical mastery increased from 53.3% to 93.3%, with a high N-Gain of 0.72 and a significant Wilcoxon test result (p = 0.000). These findings indicate that RAJA MAKAN was successfully developed through a systematic ADDIE process, offering a detective-themed, problem-oriented narrative structure that distinguishes it from existing ecosystem media, which remain largely generic and rarely embed a deliberate critical-thinking framework.