Kinesthetic Curiosity: Towards Personalized Embodied Learning with a Robot Tutor Teaching Programming in Mixed Reality

Abstract

Personalizing interactions in socially assistive robot (SAR) tutoring has shown promise with a wide variety of learners, especially when using multiple interaction modalities. Many of those interactions, however, focus on seated learning contexts, creating a need for multimodal personalization measures in kinesthetic (i.e., embodied) learning contexts. This paper proposes a multimodal measure of student kinesthetic curiosity ($KC^S$) that combines a student’s movement and curiosity measures into a single, personalized measure. This work evaluates the efficacy of $KC^S$ in a SAR tutor interaction by conducting a within-subjects ($n=9$) pilot study where participants completed kinesthetic mixed reality coding exercises alongside a curious robot tutor whose actions were determined by $KC^S$. The study results indicate that the stationarity assumptions needed for $KC^S$ were met and that the robot tutor was able to successfully use $KC^S$ to personalize its action policy, thereby positively affecting short term $KC^S$. However, no significant results were found for longer state changes for each student. The mixed reality visual programming language (MoveToCode) created for this work has been made open-source. This work aims to inform future online features and measures for mixed reality human-robot interactions.