PhD Application Portfolio · Fall 2026 · CS Education & Learning Sciences

Students leave STEM because the structure pushes them out. This site is the research program, the tools, and the curriculum I'm building to measure it and change it.

Seymour & Hunter's Talking About Leaving Revisited (2019) interviewed 5,100 STEM students across 27 institutions and found that the ones who left were not the ones who failed the coursework. Departure tracked four structural variables — teaching quality, weed-out culture, help-seeking suppression, eroded belonging — and did not track GPA or prior preparation. That finding is twenty years old at its root, it is the most cited result in STEM persistence, and it has never been replicated at a community college, where roughly half of future STEM majors begin. This site is what I would do about that.

I'm Henry Fan, a CVC-OEI Application Support Analyst at the Foothill–De Anza Community College District and a mentee of Jeff Anderson at Foothill College. I'm applying to PhD programs in CS education and the learning sciences for Fall 2026. The research agenda is four falsifiable hypotheses. The tools are three working interactive artifacts you can click. The projects are five studies with analysis plans. The methods appendix is the worked annotation, interview guide, and power calculation underneath. The curriculum is the intervention side of the same question: courses designed from the structural-equity literature, which then become research sites for it.

PhD Applicant · Fall 2026 CVC-OEI · Foothill–De Anza henry@henryfan.org


Two Practices, One Question

Most applicants separate research from curriculum design. I don't, because in my work they're the same activity: every course I design is a structural intervention, and every structural intervention is a research site I want to study. The question driving both sides is identical: what features of course design predict whether students seek help, persist, and develop belonging?

The Research

Four research questions organized around a single claim: structural features of introductory CS courses — not student ability — predict who seeks help, who persists, and who leaves. Methods: qualitative interviews (Seymour & Hunter replication), NLP analysis of course materials, learning analytics on LMS data, and computational tool-building for instructional auditing.

The Curriculum

A six-course curriculum framework for community college CS built on three principles: derive before compute, build before import, equity as design. Grounded in constructionism (Papert), productive failure (Kapur), self-determination theory (Deci & Ryan), and Jeff Anderson's antiracist learning science. Includes a signature project where students would build a working 8-bit computer on breadboards — encountering physics, linear algebra, and chemistry along the way.


Research Agenda at a Glance

Four research questions. Five empirical projects. One curriculum framework designed as the intervention side of the same question. Target venues: SIGCSE, ICER, EDM, LAK, Learning @ Scale.

Q1 · Help-Seeking What features of course design predict whether students seek help when stuck?
Learning analytics · LMS logs · discussion-forum NLP. Addressed by P1: HelpMap.
Q2 · Curriculum Structure Do dependency patterns in CS curricula predict confusion bottlenecks and DFW rates?
Curriculum graph analysis · typed dependencies. Addressed by P4: CurriculumGraph.
Q3 · Why They Left Do Seymour & Hunter's departure reasons replicate at community colleges?
Semi-structured interviews · grounded theory. Addressed by P3: Why They Left.
Q4 · Computational Auditing Can NLP match expert judgment on structural features of course materials?
Annotation study · Cohen's κ ≥ 0.65 target. Addressed by P2: SyllabusAudit and P5: BelongingSignals.

Active Projects

Qualitative Research · Interview Study · Proposed

P3 · Why They Left: A Seymour & Hunter Replication at Community Colleges

Do the structural departure reasons documented at research universities replicate at community colleges? 20–30 semi-structured interviews with students who left STEM at Foothill. IRB protocol in preparation. This project produces the conceptual foundation the other four projects need.

NLP · Instructional Design · In Progress

P2 · SyllabusAudit: NLP-Based Structural Analysis of CS Course Materials

Can automated analysis of syllabi identify structural features associated with poor student outcomes? Corpus construction and annotation schema development underway. Targeting Learning @ Scale 2027.

Curriculum Design · Constructionism · Completed

Build a Computer from Scratch: A 20-Week Cross-STEM Signature Project

Teams of community college students build a working 8-bit breadboard computer from logic gates. Seven learning science principles. Explicit STEM bridges to five disciplines. Three-track agency system. Portfolio assessment with student-proposed grades. Grounded in Papert, Kapur, Kolb, Perkins & Salomon, and Jeff Anderson's antiracist learning science. This is the curriculum that the research aims to study.

See all five research projects →


Writing

Research Statement · March 2026

Structural Predictors of Help-Seeking and Departure in Introductory CS

Four research questions, theoretical grounding, methodological approach, and why this work requires doctoral training.

Draft · available on request · henry@henryfan.org

Teaching Philosophy · March 2026

Derive Before Compute: A Teaching Philosophy for Introductory CS at Community Colleges

Constructionism, productive failure, self-determination theory, and antiracist learning science — enacted through a three-track system, portfolio assessment, and hands-on construction.

Draft · available on request · henry@henryfan.org

Full reading list with 30+ annotated entries →


Applied Work

Research-informed designs for tools and platforms that operationalize the same structural equity principles that guide the research.

ProjectBridge
A project-based learning platform for community colleges — project directory, collaborator marketplace, build journals, milestone tracking, and faculty dashboards. Designed for first-generation students. Built with Next.js, Supabase, and Tailwind. Learn more →
Interactive Tools
Three working interactive artifacts: a typed-dependency curriculum graph with live structural statistics, a pedagogical debt analyzer that scores pasted syllabus text, and a greedy MVC step-through with brute-force optimum comparison. See all three →
Teaching Computing Differently
A six-course curriculum framework for community college CS. Twelve course pages. A 20-week signature project. Open-access, no required textbooks. Visit the site →
Open Source
All papers, tools, annotation schemas, and curriculum materials released open-source. GitHub →