<![CDATA[Formative assessment in elementary science is often homogeneous, limiting its diagnostic role and support for differentiated learning. This study designs a conceptual framework for an AI-based adaptive formative assessment system to support differentiated learning in elementary science. Using R&D with the ADDIE model, the article focuses on the Analysis and Design stages. Analysis identified assessment problems and teachers' needs through semi-structured interviews and a needs questionnaire; design produced system specifications and conceptual designs from empirical findings and theory. Key outputs include: (1) an assessment blueprint with progressive difficulty, (2) adaptive logic driven by student responses, (3) a system flow diagram, and (4) an initial application prototype. It emphasizes diagnostic function, meaningful formative feedback, and teacher usability. Theoretically, it advances formative assessment research by integrating adaptivity within an assessment-for-learning perspective. In practice, the framework can guide developers, schools, and policymakers in enhancing teachers' assessment literacy and implementing adaptive formative assessment in elementary science.]]>
Copyrights © 2026
