Search

A Lightweight AI Agent for Classroom Teaching Diagnosis in Higher Vocational Education

Author(s)

Zheng Jiangwei, Qi Xinying, Wu Cuijiao

Affiliation(s)

School of Artificial Intelligence, Caofeidian Vocational and Technical College, Tangshan, Hebei, China

Abstract

Against the background of the digital transformation of vocational education, routine classroom evaluation in higher vocational colleges remains constrained by subjective judgement, fragmented criteria and weak alignment between lesson design and occupational competency requirements. Existing commercial AI systems are further limited by generic evaluation logics, dependence on video-capture infrastructure and insufficient support for integrated assessment across posts, courses, competitions and certificates. To address these limitations, this study develops XiaoQ, a lightweight school-based intelligent agent for classroom teaching diagnosis built on prompt engineering, modular role constraints and a domain-specific knowledge base. The agent supports two modes: standalone lesson-plan diagnosis and integrated diagnosis combining instructional documents with classroom observation records. We further propose a dual-track framework that separates static evaluation of instructional design from dynamic quantification of classroom behaviour, thereby distinguishing objective behavioural statistics from vocationally oriented interpretation. Application testing indicates that the agent can be deployed without dedicated hardware, has a low threshold for routine use, and produces diagnostic feedback aligned with school-based talent-training objectives. The proposed approach offers a replicable and low-cost model for data-informed teaching research and standardised classroom evaluation in higher vocational education.

Keywords

Artificial Intelligence; Classroom Teaching Diagnosis; Higher Vocational Education; Prompt Engineering; School-Based Knowledge Base

References

[1] Alon-Barkat T, et al. AI-Powered Framework for Assessing Teacher Performance in Classroom Interactions[J]. Frontiers in AI, 2025, 8: 1553051. [2] Yu H T, Guo Y Y. Generative AI Empowers Educational Reform: Current Status, Issues, and Prospects[J]. Frontiers in Education, 2023, 8: 1-15. [3] Abuhassna H. The Information Age for Education via AI and ML: A Bibliometric Analysis[J]. International Journal of Information and Education Technology, 2024, 14(5): 678-695. [4] LIU Y, ZHANG H, WANG L. A Camera-Based Visual Sensor Pipeline for Fine-Grained Human Activity Recognition in Classroom Scenes[J]. Sensors, 2026, 26(12): 3666. [5] HOU R K, FUTTERER T, BUHLER B, et al. Automated Assessment of Encouragement and Warmth in Classrooms Leveraging Multimodal Emotional Features and ChatGPT[EB/OL]. arXiv preprint arXiv:2404.15310, 2024. [6] Wang Z Q, Yang B Q, Wang S H. Enhancing Online Teaching Effectiveness Through Computer Vision Analysis[J]. Signal Image and Video Processing, 2024, 41(3): 1193-1204. [7] Wen J C, Lin Y, Si N. Behavioral Analysis of Classroom Interactions Supported by Generative AI[C]//Proceedings of the 2024 7th ICETM. ACM, 2024: 45-52. [8] Coze. Coze LLM Agent Development Platform[EB/OL]. https://www.coze.cn, 2024. [9] McCarthy B. The 4MAT System: Teaching to Learning Styles with Right/Left Mode Techniques[M]. Barrington: Excel, Inc., 1980.