In a worked example from real Erasmus+ decisions, some of CourseGraph's nearest-neighbor course mappings matched the program director's and others did not, one plausibly because it ignores differences in learning methods¶

Claim · All claims
Evidence · 1 study (2 entries) · design r2 · q2
Currency · newest study 2026 · checked against its sources 2026-10-15 (automated, 2 of 2 entries) · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly

Subclaims¶

q2 i? In the worked example for one validation student, CourseGraph's closest-match LU course corresponded with the program director's annotation for Distributed Data Systems, assigned to Distributed Systems (EDAP25). → Arthur Nijdam 2026 q2 i? CourseGraph is not fully correct: Error Correcting Codes was mapped to a basic first-year course instead of the advanced course the program director chose. → Arthur Nijdam 2026 (2)

Evidence¶

Arthur Nijdam 2026¶

Arthur Nijdam, Paul Stankovski Wagner, Sara Ramezanian. (2026). CourseGraph: Finding overlaps and differences in Computer Science courses across universities. https://arxiv.org/abs/2608.05910 doi:10.48550/arxiv.2608.05910

q2 · i? · design · r2

Qualitative hold-out analysis of 24 Erasmus+ decisions from 5 students across 3 universities (TU Delft, UC Santa Cruz, TU München), approved by the former LU program director. A t-SNE visualization (Fig. 3) shows the closest LU match "corresponds with the annotation of the program director".

"Here, we see that Distributed Data Systems gets assigned to Distributed Systems (EDAP25), which corresponds with the annotation of the program director."

Arthur Nijdam 2026 (2)¶

Arthur Nijdam, Paul Stankovski Wagner, Sara Ramezanian. (2026). CourseGraph: Finding overlaps and differences in Computer Science courses across universities. https://arxiv.org/abs/2608.05910

q2 · i? · design · r2

In the same worked example (test student 1, TU Delft), the authors give one instance of an incorrect matching: the director mapped Error Correcting Codes to the advanced EITN70, while CourseGraph chose the basic first-year EITA55. They also note that Research in Cyber Security - Hacking Lab and Systems Security were both assigned to Secure Systems Engineering (EITP20), "probably because for now, CourseGraph does not factor in differences in learning methods yet"; the Hacking Lab course is much more applied.

"Error Correcting Codes, mapped to Channel Coding for Reliable Communication (EITN70) by the program director, was mapped to Communication Systems (EITA55) by CourseGraph."

Discussion¶

Related Claims¶

• Component-level embeddings make CourseGraph interpretable: learning outcomes contributed most to an example overlap decision — related
• On the TU/e CS overlap dataset, supervised NLP classifiers (Random Forest, XGBoost) outperform thresholding and zero-shot LLM baselines at detecting course overlap — related