Diagnosing Struggles in CS/DS Education: Learning Analytics and Educational Data Mining for Data-Driven Instruction and Assessment

• In Event: Developing and Validating Tools for Student Learning and Experience

Abstract

This study investigates how knowledge component (KC) learning analytics and automatic item review can support data-driven decision-making in computer and data science education. Using student log data from an interactive online module, we identified high-difficulty and low-growth KCs and linked them with item-writing flaws using an automated rubric-based screening tool. Findings reveal persistent struggles with abstract KCs and misaligned item design, suggesting the need for instructional redesign and item revision. This work highlights the potential of integrating learning analytics and automated quality control to inform assessment improvement and instructional strategies. The study contributes to scalable, evidence-based practices in technology-enhanced learning environments and aligns with educational decision-making processes aimed at optimizing student learning outcomes.

Authors

• Jungwon Kyung, University of Massachusetts - Amherst

• Steven James Moore, Carnegie Mellon University

• Seungjoo Kim, Carnegie Mellon University