AI robots as learning companions in PBL: Effects on cognitive and self-regulatory outcomes

Abstract

This study examines the effect of robot-assisted problem-based learning (PBL) on learning outcomes in higher education. A quasi-experimental design was adopted, and the application was carried out over 16 weeks (48 hours) with two parallel branches. In the experimental group, a question-and-answer-focused educational robot was integrated into the project-based learning process, while in the control group, the same content was applied without robots. Measurements encompassed self-directed learning (including self-management), problem-solving, and comprehension/discovery dimensions. Bayesian comparisons showed that the level of evidence in favor of the experiment was in the medium-strong range for most variables; a significant increase was observed particularly in self-management. Problem-solving and comprehension/discovery scores were also higher in the experimental condition. Figures and tables report posterior effect sizes and 95% confidence intervals. The results indicate that robot-assisted PBL, which offers timely and elucidative feedback, improves the metacognitive processes associated with PBL and aids students in adhering to their study goals more consistently. In conclusion, robot-assisted PBL can significantly improve self-management, problem-solving, and comprehension outcomes, provided that appropriate feedback design and application fidelity are maintained. Future research should examine its effects across various fields by integrating long-term follow-up periods and behavioral performance assessments. The study enhances the reproducibility of results by evaluating the strength of evidence through open Bayesian reporting. This method makes design choices more transparent and establishes clear guidelines on how educational technologies should be used responsibly.