ChEdu: A Guided AI Teaching Assistant for Chemistry Education and Exam Support

作者：Xingyu Wang, Liwei Zhang, Yu Mao, Duncan J. McGillivray, Ziyun Wang · 发表于：Journal of Chemical Education · 年份：2025 · DOI：10.1021/acs.jchemed.5c00036 · 被引用次数：2 · 研究领域：Intelligent Tutoring Systems and Adaptive Learning、Science Education and Pedagogy、Educational Assessment and Pedagogy

High Resolution Image Download MS PowerPoint Slide This work introduces ChEdu, a guided AI teaching assistant addressing two critical chemistry education challenges: overwhelming demand for personalized student support and the need to foster critical thinking. Rather than introducing capabilities that LLMs fundamentally lack, ChEdu automates course-aligned guided learning in a dual-system architecture combining retrieval-augmented generation (RAG) with a fine-tuned large language model (ChEdu-GPT) grounded in Socratic questioning and zone of proximal development theory. The system’s high customizability allows academic staff to integrate their specific course materials, teaching plans, and exam resources without requiring programming expertise. ChEdu-GPT guides students through progressive questioning tailored to their knowledge levels, encouraging self-discovery rather than passive information consumption, while RAG ensures reliable retrieval of exam-related information. On a small, simulation-heavy sample (two student testers; 10 pre-exam logistics queries), retrieval was 10/10.