Mathematics Outcomes and Creative Metacognition in Primary Students: A Quasi-Experimental Study Using AI-Generated Explanations

This exploration examines the impact of supplementary AI-generated explanatory materials on foundational mathematics learning and creative metacognition in primary education. The study followed a quantitative, quasi-experimental pretest–posttest design with a nonequivalent control group. Fourth-grade classes in five schools were assigned to conditions at the classroom level: an intervention group receiving AI-facilitated handouts (n = 135) and a control group using programmatic textbooks (n = 128). Students in the experimental condition received eight supplementary lessons featuring paper-based handouts. These handouts presented explanations, examples, and illustrations of new mathematical topics generated by a large language model (Gemini 2.5 Pro), prompted the inclusion of age-appropriate content and a dedicated section encouraging creative metacognition (reflection and novel idea generation). Baseline equivalence was confirmed. The outcomes assessed included single-digit mental arithmetic, multidigit calculation, arithmetic word problem-solving, and creative metacognition, which were analyzed via multilevel or robust linear modeling accounting for classroom clustering. The results indicated no statistically significant differences between groups on single-digit mental arithmetic tasks. However, students who received AI-generated materials showed significantly greater improvement in multidigit calculation tasks and one-step word problem-solving than did the control group. No significant difference was found for multistep word problems. The intervention group also presented a substantially greater increase in creative metacognition scores than did the control group. The findings provide preliminary quasi-experimental evidence that curated, static AI-generated explanations, when designed to include metacognitive prompts, may be associated with greater gains in specific procedural and applied mathematical skills and creative metacognition among fourth-grade students under the conditions studied. This approach illustrates how generative AI can be leveraged as an authoring tool to supplement standard instruction and cultivate higher-order cognitive skills in young learners.