Development and Evaluation of an AI-Assisted Robotic Game-Based Learning System for Enhancing Logical Reasoning

Promoting learners’ knowledge construction during inquiry-based learning has long been a central concern in the learning sciences. To support learners’ logical reasoning, this study developed an instructional system integrating generative AI-based dynamic scaffolding with a physical robotic arm. The system incorporates two logic reasoning tasks—the Bottle-Taking Game and the River-Crossing Game—to provide adaptive AI-assisted guidance through embodied interaction for problem-solving. A quasi-experimental nonequivalent pretest–posttest design was conducted with 64 sixth-grade students. The experimental group (n = 31) learned through an AI-assisted robotic interaction mode, whereas the control group (n = 33) received conventional hands-on instruction. The results showed that the experimental group achieved significantly better logical reasoning performance than the control group. ARCS-V motivation analysis revealed that high-achieving students in the experimental group scored significantly higher on the Attention, Relevance, Confidence, and Volition dimensions than their peers in the control group, whereas low-achieving students scored significantly lower on Relevance. No significant between-group differences were found in cognitive load, indicating that AI-assisted scaffolding did not impose additional cognitive burden. Furthermore, students perceived the proposed system as useful, easy to use, and satisfactory. Overall, the findings demonstrate that integrating AI-assisted scaffolding with embodied robotic interaction enhances logical reasoning performance while maintaining learning motivation and cognitive processing and fostering positive technology acceptance. These results provide empirical evidence supporting AI-assisted robotic game-based learning and contribute to STEM education.