Abstract: This research examines how adding an AI tool “MagicSchool” to a makerspace STEM program affects elementary students' ability to solve problems. Three cognitive subdomains—strategic planning, creative ideation, and iterative perseverance—were examined using a true experimental, post-test-only control-group design with 60 sixth graders, ages 11–12, from two public schools in Jerusalem. Students were randomly assigned equally to the experimental group, AI-supported makerspace, and the control group, makerspace-only. For 12 sessions, both groups completed 15 engineering challenges. The effects of the intervention were assessed after its completion by employing the validated Elementary Problem-Solving Inventory, EPSI. Statistical analysis involved a multivariate analysis of variance, MANOVA, which was subsequently followed by univariate analyses of variance, ANOVAs, with effect sizes also being reported. MANOVA resulted in a significant multivariate effect of instructional condition on problem-solving variables: $$ \text{Wilks' } \Lambda = 0.624,\quad F(3,56)=11.26,\quad p<.001,\quad \eta_p^2=0.376 $$ The ANOVAs indicated substantial improvements for the AI group in strategic planning: $$ p<.001,\quad \eta_p^2=0.198 $$ creative ideation: $$ p=.003,\quad \eta_p^2=0.143 $$ and iterative perseverance: $$ p=.011,\quad \eta_p^2=0.106 $$ Integrating MagicSchool AI significantly improved elementary students’ problem-solving competence across all measured domains. These findings offer causal evidence supporting AI scaffolding in hands-on, constructionist learning environments and inform curriculum and teacher support in primary STEM education.

Keywords: MagicSchool AI, makerspace, problem-solving, strategic planning, creative thinking, iterative learning, stem education, elementary school, experimental design

Pages: 26-34