Earthquaker-AI: RAG Framework for Primary School Earthquake Education

What changed

This paper introduces Earthquaker-AI, an educational framework that enhances a prior robotics project by incorporating a conversational AI assistant powered by Retrieval-Augmented Generation (RAG). The system is designed to improve earthquake preparedness and encourage conscious action among primary school students. It builds upon the existing "Earthquaker" STEM project, which previously used Lego WeDo2 robotics for mechanical simulation of seismic responses. Earthquaker-AI extends this by focusing on cognitive and metacognitive processing, moving beyond tangible sensor and actuator interactions to include AI-driven learning.

The AI assistant functions as a guided learning mechanism. It aligns student responses with established safety guidelines and provides verbal feedback based on rubrics. This feedback aims to foster self-regulated learning and promote calmness in simulated emergency scenarios. The framework adopts a progressive learning trajectory tailored to cognitive development:

• Early Grades: Focuses on basic recognition of safety actions through multiple-choice questions, assessed using a two-dimensional rubric.
• Middle Grades: Students identify correct action sequences via multiple-choice questions, evaluated with a three-axis rubric.
• Upper Grades: The approach shifts to verbal production, requiring short written responses that are assessed by a four-dimensional rubric, including clarity of expression.

The core dialogic module utilizes RAG to semantically match student queries with official safety guidelines, generating responses that are intended to be safe and accurate. Experimental evaluations reported in the paper indicate high groundedness and accuracy, alongside a low hallucination rate.

Why it matters for builders

Earthquaker-AI presents an innovative model for educational technology, demonstrating how RAG can be effectively applied to deliver safety-critical information in an engaging and adaptive manner. Builders interested in educational AI or RAG applications can draw inspiration from its multi-stage assessment approach and its integration with physical computing elements. The framework highlights the potential for AI to support metacognitive skills and emotional regulation in learning environments, particularly for complex or high-stakes topics like disaster preparedness.

Practical impact

Developers can explore implementing similar RAG-based conversational agents for educational purposes, focusing on domain-specific knowledge bases and structured assessment rubrics. The progressive learning design suggests a pathway for creating adaptive learning systems that cater to different age groups and cognitive levels. The combination of robotics and AI offers a blueprint for interactive STEM learning experiences that promote not only technical understanding but also critical thinking and responsible behavior. Builders might consider adapting the rubric-based feedback mechanism for other AI tutoring systems to enhance student self-assessment and learning outcomes.

Caveats and source limits

The primary source for this information is a research paper published on arXiv. While the paper reports experimental evaluation showing high groundedness and accuracy with a low hallucination rate, these results are based on the authors' own evaluation. Independent benchmark results or third-party validation of Earthquaker-AI's effectiveness are not provided. The specific details of the RAG implementation, including the exact models, retrieval mechanisms, and the size/nature of the knowledge base used, are not elaborated upon in the excerpt. Furthermore, the practical deployment challenges and scalability of such a hybrid system are not discussed.