<![CDATA[The use of generative artificial intelligence in primary education is rapidly expanding; however, empirical evidence regarding its effectiveness in improving primary schools mathematics learning outcomes remains limited. This study aims to compare the effectiveness of ChatGPT-assisted independent practice with conventional teacher-guided practice in elementary mathematics learning. Using a quasi-experimental design, two groups of students received different treatments across four practice sessions covering decimal operations and comparison concepts. The findings reveal that students who practiced with ChatGPT achieved significantly higher learning outcomes than those in the conventional group, with a very large effect size (Cohen’s d = 1.39). These results align with self-regulated learning theory, which highlights the importance of adaptive scaffolding in enhancing the quality of independent practice. ChatGPT provides instant explanations, just-in-time feedback, and corrective examples that directly address misconceptions—advantages not typically available in conventional practice. This study underscores the potential of ChatGPT as a learning companion in elementary mathematics and opens avenues for further research on the design of safe, structured, and developmentally appropriate AI–learner interactions.]]>
Copyrights © 2025
