Curriculum Alignment: Measuring CS2013 vs CS2023 Coverage

Sherzod Turaev and five coauthors submitted a paper on 17 Jun 2026 that applies a human-in-the-loop pipeline to measure how an accredited BSc in Computer Science aligns with Computer Science Curricula 2013 (CS2013) and 2023 (CS2023). The study finds the program covers 49.7% of CS2023 knowledge units and 50.9% of CS2013 knowledge units, with notable differences in delivered cognitive depth.

What did the study measure and how?

The paper measures topical coverage, competency articulation, and cognitive depth by representing the program and each guideline as structured corpora, generating candidate course-to-knowledge-unit matches with semantic retrieval, and confirming matches through human judgment under an explicit coverage definition. The workflow is a retrieve-then-confirm pipeline that the authors validate with a second rater and benchmark seven retrievers, finding a reciprocal-rank-fusion ensemble to be the strongest retriever while a reputed long-context model underperformed a small sentence model.

What did the audit find?

The audit reports that the program covers 49.7% of CS2023 knowledge units and 50.9% of CS2013 units, a near-constant share across the decade. Of the units the program covers under either guideline, the program articulates the competency for approximately 88% of those covered units, yet it delivers competency at the recommended cognitive depth for 76% of present units under CS2023 versus 95% under CS2013. The longitudinal comparison isolates persistent structural gaps that appear against both guidelines and ABET: parallel and distributed computing, foundations of programming languages, and systems fundamentals. Inter-rater validation returned Cohen's kappa values of 0.64 for CS2023 and 0.69 for CS2013.

Why it matters

The numbers show that mere topical coverage stayed almost unchanged across a decade, while the newer guideline raised depth expectations that the program meets less often. That gap is not evidence the program declined, the authors say, but that CS2023 raises recommended cognitive depth relative to CS2013. The finding separates two policy levers: programs can increase unit coverage, or they can raise instructional depth where competencies are already present.

What to watch

Watch whether programs responding to CS2023 shift resources from breadth to deeper treatment in areas flagged as persistent gaps, notably parallel and distributed computing, programming-languages foundations, and systems fundamentals. Also watch method choices: the paper shows retriever selection matters, with a reciprocal-rank-fusion ensemble outperforming a long-context model in this task.

The authors note the instrument is reusable and available on request, and that their maps were validated with an independent second rater. The paper includes five figures and eight tables documenting the pipeline, retriever benchmarks, validation, and the longitudinal comparisons.