Beyond Single Scores: A Dual-Dimensional Assessment of Student Mastery in AI

• In Event: Virtual Poster Session
  In Virtual iPresentation Only: SIG-Technology, Instruction, Cognition & Learning Virtual Poster Session

Abstract

Precise student assessment is crucial in AI talent cultivation. While Deep Knowledge Tracing (DKT) effectively predicts knowledge mastery, its results lack the interpretability for actionable feedback. Addressing this challenge in AI education, our study proposes an integrated frameworkthat constructs an AI discipline knowledge graph and uses Bloom’s Taxonomy to guide structured assessment. The core of the framework is its dual-dimensional interpretation of DKT's diagnostic results via cognitive level and knowledge structure. We validated the framework by analyzing undergraduate AI students' assessment data with a DKT model. Experiments Results show that the framework can precisely identify individual cognitive profiles and reveal deep cognitive differences among students, guiding precision pedagogy and personalized support in higher education’s AI major.

Authors

• Zhen Li, Central China Normal University

• Chenxuan He, Central China Normal University

• Shangheng Du, East China Normal University