Robot Quiz Study Links Algorithm Accuracy to Student Motivation

A new study from Rezaul Tutul, Ilona Buchem, and Niels Pinkwart (Humboldt University) investigates how the accuracy of sound detection algorithms influences student motivation during a robot-mediated quiz game. Using a Pepper humanoid robot as host, the researchers built an interactive buzzer-based quiz and compared two approaches for identifying the first responder: a Convolutional Neural Network (CNN) and a classic Cross-Correlation algorithm. In a controlled between-subjects experiment with 40 university students (20 per group), both groups completed identical quizzes under the same conditions, differing only in the detection algorithm. Student motivation was measured using the Intrinsic Motivation Inventory (IMI), while real-time detection accuracy was logged.

The results show that the Cross-Correlation approach achieved more reliable sound detection in real classroom conditions, despite being simpler than the deep learning-based CNN. Crucially, this accuracy premium translated directly into higher motivation: Cross-Correlation group participants scored significantly higher on all IMI subscales—greater interest/enjoyment, perceived competence, effort, perceived choice, and lower pressure (reverse-coded). The authors introduce the Algorithmic Precision-Motivation Relationship (APMR) model, empirically demonstrating that algorithmic accuracy is not just an engineering metric but a psychological factor that shapes learners' sense of fairness, competence, and overall engagement in robot-assisted education. This suggests that developers should prioritize reliability over algorithmic complexity when designing interactive educational robots.