Abstract

This study proposes and validates the system architecture of an Intelligent Supervision Support System for Vocational Dual Education (ISS-DVE), developed to address critical limitations in traditional supervision practices, including fragmented data management, inconsistent advisory processes, and limited utilization of feedback for continuous improvement. The proposed conceptual framework integrates the Input–Process–Output–Feedback (IPOF) system model with an Artificial Intelligence Agent architecture based on the Perceive–Decide–Act–Learn (PDAL) cycle, together with Retrieval-Augmented Generation combined with Large Language Models (RAG+LLM) to enhance contextual reasoning, evidence-grounded recommendations, and transparency in supervisory decision-making. The appropriateness of the conceptual framework was evaluated through expert validation involving 15 specialists across system and technical, artificial intelligence and decision-making, and user and quality assurance domains. Using a five-point Likert scale, data were analyzed by mean, standard deviation, and the Index of Item-Objective Congruence (IOC). The evaluation results demonstrated an excellent level of overall appropriateness (mean = 4.64, S.D. = 0.37) and strong content validity (IOC = 0.83), with the System dimension receiving the highest rating (mean = 4.87, S.D. = 0.25), followed by User & Environment, Decision, Learning, and Intelligence dimensions. These findings indicate that the ISS-DVE architecture is theoretically sound, structurally coherent, and suitable for further development into a prototype system to support intelligent, transparent, and adaptive supervision in vocational dual education contexts.

This work is licensed under a Creative Commons Attribution 4.0 License.