A Knowledge-Driven Multi-Evaluation Framework for Art Pedagogy in the Digital Era

Abstract

Traditional art teaching evaluation, relying on subjective judgment and uniform criteria, fails to support individualized learning. Grounded in knowledge management principles, this study develops a knowledge-driven, multi-agent framework for art teaching evaluation that integrates multiple data sources, diagnostic, formative, and summative assessment phases, and a closed-loop feedback mechanism with dynamic weight adjustment, converting behavioral data analytics into actionable pedagogical intelligence. A quasi-experimental comparison of 186 students across two art courses shows that the framework significantly enhances creativity and engagement relative to traditional score-based evaluation, while fostering collaborative knowledge sharing among instructors. Implementation nonetheless faces data-integration and platform-reliability challenges that increase instructors' workload. Future research will strengthen the platform's anomaly-detection and data-repair capabilities and further validate the framework's bias-correction mechanisms and generalizability.