Research
My active research program is three first-author papers in applied linear algebra and CS education, in progress with Prof. Jeff Anderson (Foothill College) — tracked openly on the Modeling Bench. Alongside that work, I am developing a broader research agenda in community-college CS education — four questions about dependency structure, help-seeking, identity, and algorithmic equity — that I intend to formalize in a PhD. This page separates the two clearly: what’s in flight, and what I want to pursue next.
Positioning
Two things are running in parallel. The first is concrete and near-term: a three-paper program anchored to Prof. Jeff Anderson’s published work in applied linear algebra and his anti-racist, learner-centered objectives. Those drafts are active, have target venues, and are tracked week by week on the Modeling Bench.
The second is a broader research agenda that emerged from designing the proposed community-college CS curriculum on this site and from my day job at CVC-OEI — the California Community Colleges’ cross-enrollment infrastructure, serving 115+ colleges. Four research questions about dependency structure, help-seeking, identity formation, and algorithmic equity — not imported from a literature review but drawn from what I see every working day about who stays in CS and who leaves.
Community colleges serve nearly half of all undergraduates in the United States, yet most CS-education research is conducted at four-year residential institutions with very different student populations. The CS-Education agenda below is the work I want to formalize in a PhD. It is a proposed program of research, not a record of completed studies.
Three first-author papers in progress with Prof. Jeff Anderson, my mentor in applied linear algebra at Foothill College. Each paper extends one of Jeff’s published projects or fills in one he marked “under development” in his Math 2BL deliverables. This is the concrete work: draft abstracts, weekly plans, named risks, target venues.
The Modeling
Bench
MER 2.0
Open-Hardware Lab Kit for Coupled-Oscillator Modeling
PRIMUSMatrices → Networks
Linear-Algebra-First NNs via SVD of Trained Layers
PRIMUSAI Modeling Tutor
Anti-Racist Audit Framework for LLM Tutors
JCHEMentor: Prof. Jeff Anderson · Foothill College · Mirror: fansofhenry.github.io/research-lab
PhD Research Questions
Four interconnected questions about equity, structure, and learning in community-college CS — each grounded in curriculum design practice and connected to a proposed intervention. These are the questions I want to formalize in a doctoral program, not studies I have already run.
How do curriculum dependency structures in introductory CS courses shape who persists and who leaves — and how can those structures be redesigned to broaden participation?
Community college CS students face prerequisite chains that were designed for a different population. When a student fails CS1, the structure forces a full-year delay. This question investigates whether the dependency graph itself — not student ability — is the primary attrition mechanism.
The three-track system proposed in every course replaces prerequisite gates with depth choices, allowing students to enter any course at Track I without prior programming experience.
What are the help-seeking behaviors of community college CS students, and how do structural barriers (work schedules, commute times, family obligations) reshape when and how students seek help?
Help-seeking research in CS education overwhelmingly studies four-year residential students. Community college students — who are more likely to work full-time, commute, and have caregiving responsibilities — have fundamentally different access to office hours, study groups, and peer networks. Their help-seeking patterns are not deficits; they are rational responses to structural constraints.
Every course in the proposal is designed with asynchronous help channels, recorded walkthroughs of common stuck points, and project milestones sequenced so that the most common failure modes are addressable without synchronous faculty contact.
How does a project-based, no-exam assessment model affect CS identity formation among students who have been historically excluded from computing?
Students who do not see themselves as 'CS people' often cite exam culture, competitive grading, and the assumption of prior exposure as the reasons. Portfolio-based assessment removes the high-stakes single-point-of-failure that exams represent. This question asks whether that structural change affects whether students begin to identify as people who do computing.
The portfolio assessment model, public exhibitions, and student-proposed grading specified across all six courses are the proposed intervention. The Abuelita Test — which asks students to explain their work to a non-technical family member — is designed specifically to strengthen identity through communication.
How do community college students experience and make meaning of algorithmic bias — and does integrating equity analysis into technical coursework change how they understand the systems they build?
Most research on ethics in CS education studies elite four-year institutions. Community college students bring different lived experiences with algorithmic systems — as subjects of automated hiring tools, predictive policing, and benefits allocation algorithms. Their analysis of bias is not abstract; it is personal. This question investigates how that positionality shapes critical technical practice.
The Bias Audit project designed for CS 180 and CS 185, modeled on Buolamwini's Gender Shades methodology, is the proposed site for this research. In the course design, students choose real deployed systems, design test protocols, measure performance across demographic groups, and write findings reports.
Proposed Project Designs
Five study designs I would bring to a doctoral program, each tied to one of the research questions above. These are proposed research designs — methodology, data, venue, and expected contribution — not data collection that has already happened.
All five are pre-IRB, pre-data. Status is planning, not in-flight.
HelpMap
ProposedStructural Predictors of Help-Seeking in Introductory CS
LMS learning analytics to identify structural predictors of help-seeking behavior. Students who need help most are least likely to seek it — not because of individual deficit but because of course design. This project builds a predictive model using LMS interaction data to identify students at risk of help-seeking suppression.
Logistic regression on LMS interaction logs. Feature families: temporal (session timing, spacing), linguistic (forum post complexity, question specificity), engagement (resource access patterns, assignment submission timing). Target: help-seeking events (office hours, tutoring center, forum posts).
SyllabusAudit
ProposedNLP Classification of Pedagogical Debt in CS Syllabi
An NLP classifier that detects 'pedagogical debt' — violations of Harel's intellectual need principle — in CS course syllabi. If a syllabus introduces hash tables in Week 3 without a motivating problem, that's pedagogical debt: abstraction before need. This tool makes that debt measurable.
47-syllabus corpus from California community colleges. Annotation schema with 4 dimensions: necessity (is the concept motivated?), sequence (does prerequisite precede dependent?), equity (are multiple entry points offered?), assessment (is assessment aligned to stated learning goals?). Cohen's kappa target ≥ 0.7.
Why They Left
ProposedDeparture Narratives from Community College CS
20–30 semi-structured interviews with students who left CS at community colleges. Not 'why did you fail?' but 'what was your experience, and what would have changed it?' Using Seymour & Hunter's taxonomy of STEM departure factors, adapted for community college context.
Qualitative: semi-structured interviews, 60–90 minutes. IRB protocol. Deductive coding (Seymour & Hunter departure taxonomy) + inductive coding (grounded theory for emergent themes). Member checking with participants.
CurriculumGraph
ProposedGraph-Theoretic Analysis of CS Curriculum Dependencies
A tool that models CS curricula as directed graphs with typed dependencies. Four dependency types: conceptual (topic A requires understanding topic B), procedural (skill A requires practicing skill B first), motivational (topic A provides intellectual need for topic B), and social (activity A builds trust needed for activity B). Structural graph properties predict student outcomes.
Model courses as directed acyclic graphs. Node = concept/skill. Edge = typed dependency. Compute: longest path (curriculum bottleneck), in-degree distribution (prerequisite load), connected components (isolated topics). Pilot with 5–10 instructors for validation.
BelongingSignals
ProposedMeasuring Belonging Cues in CS Course Materials
A coding scheme for identifying belonging signals (and threats) in CS course materials — syllabi, assignments, code comments, error messages, course websites. Based on Walton's belonging uncertainty framework: whether students believe 'people like me' are expected to succeed in CS.
Walton 3-item belonging scale administered pre/post. Coding scheme applied to course materials. Instructor self-audit instrument developed for voluntary use. Correlation between material-level belonging signals and student-reported belonging.
My methodological orientation is mixed-methods with a qualitative core. The questions I ask require understanding student experience from the inside — not just measuring outcomes at scale.
Design-Based Research (DBR)
Iterative cycles of design, implementation, analysis, and redesign — conducted in authentic educational settings. The curriculum itself is the intervention, and each semester is a research cycle.
Qualitative Methods
Semi-structured interviews, think-aloud protocols, artifact analysis, and thematic coding. Understanding the student experience requires methods that center student voice.
Learning Analytics
Curriculum network analysis, prerequisite dependency modeling, and survival analysis applied to institutional data to understand structural barriers at scale.
Participatory Design
Students as co-designers of curriculum, not just subjects of study. Proposed three-track system to be developed with student input across multiple semesters of iteration.
PhD Program Interests
I am looking for doctoral programs where curriculum design and research are not separate activities — where the work of designing equitable computing education is itself a form of scholarly inquiry.
CS Education
Computing education research, broadening participation, curriculum design
University of Washington (Ko), University of Michigan, Georgia Tech, UC San Diego
Learning Sciences
How people learn in authentic contexts, design-based research, equity in STEM
Northwestern, Stanford, University of Colorado Boulder, Vanderbilt
HCI / Social Computing
Human-computer interaction, algorithmic fairness, participatory design
CMU HCII, University of Washington, MIT Media Lab, Cornell IS
“The curriculum is the proposal. The Modeling Bench is the proof that the proposal is being built in the open, with a mentor, on a clock.”
Research · Practice Connection