Measuring Curriculum Alignment across Topical Coverage, Competency, and Cognitive Depth: A Longitudinal Framework Applied to CS2013 and CS2023

Researchers developed a human-in-the-loop pipeline to measure how well computer science undergraduate programs align with international curricular guidelines, applying it longitudinally to CS2013 and CS2023 standards. The analysis reveals persistent structural gaps in parallel computing, programming languages, and systems fundamentals across both decades, while showing program coverage remained near-constant at ~50% despite guideline restructuring.

This research addresses a critical gap in computer science education accountability by establishing the first reproducible framework for measuring curriculum-guideline alignment at scale. Universities have lacked standardized tools to benchmark their offerings against evolving international standards like those from ACM/IEEE, making it difficult to identify systematic coverage gaps or track improvements over time. The study's longitudinal application to a single accredited program reveals that despite a decade of guideline revisions, the program's overall knowledge-unit coverage remained essentially flat at approximately 50%, suggesting either institutional inertia or fundamental constraints in curriculum design that warrant investigation.

The research methodology combines semantic retrieval with human validation, finding that ensemble approaches outperform individual retrievers and that smaller, specialized models can exceed large language models in specialized domains. This has implications for educational technology development and suggests that off-the-shelf AI solutions require careful benchmarking before deployment in academic contexts.

The findings identify three persistent structural gaps—parallel computing, programming language foundations, and systems fundamentals—that remain uncovered across both CS2013 and CS2023 standards and ABET accreditation requirements. These gaps likely reflect resource constraints, faculty expertise limitations, or curricular prioritization rather than guideline ambiguity. The disparity in cognitive depth delivery between guidelines (76% for CS2023 versus 95% for CS2013) indicates that newer standards have raised expectations faster than programs can adapt their teaching approaches.

The tool's reusability and explicit validation methodology could enable broader comparative studies across institutions, potentially driving systemic improvements in computer science education quality and consistency worldwide.

• →A new measurement framework reveals computer science programs cover approximately 50% of international guideline knowledge units with minimal change across a decade
• →Persistent gaps in parallel computing, programming languages, and systems fundamentals exist across both CS2013 and CS2023 standards, suggesting systemic curriculum design challenges
• →Semantic retrieval ensemble methods outperform individual large language models for educational domain matching, requiring careful benchmarking before deployment
• →Programs deliver recommended cognitive depth for 95% of CS2013 units but only 76% of CS2023 units, reflecting raised expectations in newer guidelines
• →Human-validated, longitudinal curriculum assessment enables separation of institutional gaps from standard evolution