Abstract
Abstract This study explores the potential and integration of a Swahili-speaking social robot as a mathematics tutor in Tanzanian primary schools. Leveraging the GPT-3.5 Large Language Model (LLM) and a NAO social robot, the research investigates the feasibility and effectiveness of using state-of-the-art technologies to support mathematics education in a low-resource setting. Because classroom lessons in such contexts often provide limited opportunities for individualised support, the robot was designed to serve as an after-class remedial tutor following a lesson on the same topic taught by a human teacher. Through a series of sessions conducted in five public primary schools involving 26 third- and fourth-grade pupils, the study assessed pupils’ cognitive learning outcomes and their perceptions of the social robot as a tutor. The robot’s tutoring approach incorporated contextual information and pupils’ progress to provide personalised feedback. The results show a significant improvement in pupils’ understanding of the mathematics topic following the robot’s tutoring, supported by qualitative feedback highlighting the robot’s friendliness, knowledgeability, and effective tutoring methods. The study underscores the potential of AI-powered social robots to complement classroom teaching in resource-constrained environments and offers insights for future research and development in this domain.
Showing the abstract — retrieve the full paper via the Exa API.