<![CDATA[Artificial intelligence (AI) is increasingly used to support adaptive feedback, prompts, and learning analytics in science education, yet the mechanisms linking AI-supported scaffolding to conceptual understanding remain underexplored. This study examined whether metacognitive awareness mediated the association between AI-supported cognitive scaffolding and conceptual understanding, and whether prior knowledge moderated the scaffolding–metacognition pathway. A quasi-experimental pretest–posttest control-group design was conducted with 214 Grade 9–10 students from four intact science classes in Bandar Lampung, Indonesia. Students were assigned to either an AI-scaffolding condition (n = 108) or a conventional instruction condition (n = 106) over a 12-week intervention. Conceptual understanding was assessed using a 40-item curriculum-aligned test (α = .87), while metacognitive awareness was measured using a 30-item adapted inventory (α = .82). Results showed that students in the AI-scaffolding condition achievedhigher post-test scores and greater learning gains. Metacognitive awareness partially mediated this association, and the indirect effect was stronger among students with lower prior knowledge.]]>
Copyrights © 2026
