Peer-reviewed CS education research.

• ResearcherYusuf Gadelrab — BS Computer Science, San José State University, expected May 2028, dual track in applied mathematics and computer science
• LabCS Education Research Lab, San José State University
• AdvisorDr. Ethel Tshukudu
• VenueACM SIGCSE Technical Symposium 2026 (SIGCSE TS 2026)
• MethodsIRB-approved protocol, mixed methods, 60 participants
• Primary DOI10.1145/3770761.3777339
• Contactyusuf.gadelrab06@gmail.com · GitHub · LinkedIn

Programming languages are written in English. if, while, return, every error message, every tutorial, every Stack Overflow answer. For a student who is still learning English, that means the first CS class is really two classes taken at once — one in computing and one in a second language — and only one of them is being taught. The dropout that follows gets read as "not a coder." It is often just load.

The first paper tests the obvious intervention nobody had measured carefully at this level: let novices meet the material in their first language alongside the English syntax, and see what happens to how they perform and, just as importantly, to whether they believe they belong in the room. The second paper works the other end of the same problem. If a curriculum could see what a specific learner already knows and re-sequence itself around that, the language gap stops being a wall and becomes a variable the system can plan around.

Both papers sit in computer science education research, a field whose job is to replace intuitions about who "gets" programming with measurements.

1. Exploring Bilingual Coding for Inclusive CS Learning

   SIGCSE Technical Symposium 2026 · ACM · DOI 10.1145/3770761.3777339

   A mixed-methods, IRB-approved study of 60 participants asking whether novices learn programming better when they can read and write code alongside their first language, rather than being forced through English-only syntax and English-only instruction from day one.

   The design pairs quantitative performance data with qualitative accounts from the learners themselves, because the interesting failure mode in introductory CS is rarely a wrong answer — it is a student who decides, quietly and early, that this subject is not for them. The headline result: statistically significant pre-to-post gains in confidence, computing identity, enjoyment and motivation, with inexperienced participants gaining significantly more in confidence than experienced ones. Confidence is the measure that predicts whether someone takes the second CS course, which is the outcome the field actually cares about.

   Read it at doi.org/10.1145/3770761.3777339 →

   Gadelrab, Y. et al. (2026). Exploring Bilingual Coding for Inclusive CS Learning. In Proceedings of the 57th ACM Technical Symposium on Computer Science Education (SIGCSE TS 2026). ACM. https://doi.org/10.1145/3770761.3777339

2. Adaptive Curriculum Maps: Graph-Augmented Retrieval-Oriented LLMs for Education

   SIGCSE Technical Symposium 2026 · ACM · poster

   Combining knowledge graphs with retrieval-augmented large language models to generate computer science curricula that adapt to what a learner already knows, instead of restarting every learner at the same lesson one.

   A knowledge graph gives the model the prerequisite structure of a subject — what depends on what — so retrieval is constrained by the shape of the discipline rather than by whatever the embedding happened to find nearby. That is what separates an adaptive curriculum map from a chatbot that can talk about loops: the graph is what stops the system from confidently teaching recursion to someone who has not yet met a function.

   Gadelrab, Y. et al. (2026). Adaptive Curriculum Maps: Graph-Augmented Retrieval-Oriented LLMs for Education. Poster, 57th ACM Technical Symposium on Computer Science Education (SIGCSE TS 2026). ACM.

The peer-reviewed paper is the source of record. Anything reported here beyond what is stated above should be read out of the paper itself, not out of this page.