Earthquaker-AI Enhances Primary School Earthquake Education with RAG and Robotics

This paper introduces Earthquaker-AI, an innovative educational framework designed to improve earthquake preparedness and response skills among primary school students. Building on an existing robotics project, it integrates a conversational AI assistant powered by Retrieval-Augmented Generation (RAG). The system aims to move beyond mechanical simulation to foster cognitive and metacognitive processing in young learners. The framework utilizes Lego WeDo2 robotics for hands-on simulation of seismic responses, allowing students to interact with sensors and actuators. The AI assistant acts as a guided learning tool, aligning student responses with official safety guidelines and providing structured, rubric-based verbal feedback. This feedback supports self-regulated learning and helps students maintain calmness during emergencies. The learning trajectory is progressive, adapting assessment rubrics and question types (multiple-choice to short written responses) to suit different primary grade levels. Experimental evaluations indicate that Earthquaker-AI achieves high groundedness and accuracy, with a low rate of AI hallucination. By combining robotics, structured rubrics, and AI, the system promotes technological literacy, self-regulation, and responsible digital use, while also developing crucial early crisis-management skills.