Deep Learning Pedagogies Enhance AI Literacy in Elementary Students: A Five-Cycle Implementation Study

The urgency to integrate Artificial Intelligence (AI) literacy into primary education in Indonesia is driven by the increasing presence of AI technologies in everyday life and the nation’s strategic vision to prepare a future-ready workforce. However, current teaching practices remain largely behavioristic and content-driven, lacking the pedagogical depth needed to foster conceptual understanding and ethical engagement with AI. This study addresses this gap by investigating how deep-learning pedagogies—approaches that pursue deep learning as a goal through active, reflective, and collaborative experiences— can be used to improve AI literacy among fifth-grade students. Grounded in design-based research (DBR), the study implemented and refined the Associative Model of AI Literacy (AMAIL), a framework integrating cognitive constructivism, social constructivism, constructionism, and transformative learning theories. The intervention spanned five cycles in three public schools in Salatiga, involving 118 students. Learning outcomes were assessed using pre- and post-tests and student reflections, with analysis conducted through bootstrap methods and Exact McNemar’s tests. Findings showed statistically significant improvements in students’ ability to recognize AI, explain its logic, and reflect on its ethical implications. The study demonstrates how deep-learning approaches, when applied iteratively and contextually, can foster not only technical understanding but also critical and ethical AI literacy in primary education. These findings can inform educators and government stakeholders in designing and implementing pedagogical strategies that support comprehensive AI literacy development at the primary level