Overview

The complexity of modern software engineering (SE) curricula increasingly outpaces traditional lecture-based instruction, leaving students underserved by generic supplementary materials. My research introduces context-aware, real-time adaptive personalization of supplementary resources using lightweight, interpretable algorithms (explicitly avoiding the biases, costs, and opacity of AI-generated content). Anchored in my PhD dissertation, Learning Beyond the Lecture: Rethinking Software Engineering Education with Supplementary Materials in the Generative AI Era, I develop evidence-based systems that optimize student engagement and comprehension in SE courses. My vision is to establish a scalable, preference-driven framework for adaptive learning support that originates in SE and extends to computing education research (CER), transforming how technical concepts are reinforced in diverse classroom settings.

Dissertation Summary

My dissertation confronted the limitations of static or randomized supplementary materials in SE education, where individual differences in preferences and context lead to inconsistent learning outcomes. Most adaptive learning platforms (ALPs) rely on generative AI and open educational resources (OER), introducing algorithmic bias, high implementation costs, and reduced instructor control. I proposed contextual Thompson Sampling (TS), a bandit algorithm that personalizes material selection using live student ratings from a curated, high-quality resource pool, requiring no AI content generation.

Empirical Validation

I validated this approach through three rigorous classroom-based studies:

• Study 1 — TS vs Randomization: Compared TS-driven adaptation to random selection in an SE course, showing superior median usefulness and improved positive/negative rating ratios across topics and quizzes.
• Study 2 — Personalization Factors: Examined learner preferences, visual/verbal tendencies, and gender via adaptive experiments with IRB-approved protocols. Learner preferences were the strongest predictor of perceived usefulness; relevance, clarity, detail, and examples emerged as key mediators. Visual learners rated personalized materials higher when aligned with preferences; verbal learners assigned lower ratings overall. Aligning delivery mode by learning style had minimal effect on perceived usefulness, reinforcing critiques of strict learning-style matching. Gender showed no significant effects.
• Study 3 — Student Preferences Survey: A comprehensive survey of SE students identified nuanced preferences for material design (e.g., concise examples, structured explanations), which directly informed algorithm refinement. All studies included thorough validity assessments (internal, construct, external).

Current Platform Work

I am developing a digital platform that deploys contextual TS to deliver real-time, preference-aligned supplementary materials in SE courses. The system uses live student feedback (not AI) to recommend from vetted resources, addressing common criticisms of ALPs. Generative AI is used only for ancillary tasks—producing concise concept summaries and quiz items to support assessment—not for creating core supplementary content.

Research Phases

I will pursue a phased research program as a faculty member:

• Phase 1 (SE Deepening): Longitudinal studies in advanced SE topics to measure retention, project performance, and long-term preference evolution.
• Phase 2 (CER Expansion): Adapt the TS framework to foundational computing courses (e.g., programming, algorithms) to test generalizability across subdisciplines.
• Phase 3 (Open-Source Ecosystem): Develop and release a modular, instructor-friendly ALP toolkit enabling contextual TS deployment with custom material libraries, prioritizing accessibility for under-resourced programs.

Impact, Funding, and Mentorship

This agenda aligns with Computer Science Department priorities in computing education research and human-centered systems. I plan to pursue NSF and industry funding, mentor students in mixed-methods education research, and integrate adaptive tools into SE pedagogy.

My research reimagines supplementary learning in SE education through transparent, preference-driven, real-time personalization—offering a robust, scalable alternative to AI-heavy platforms. By joining the department, I will advance a sustainable, inclusive model for computing education that empowers students, equips instructors, and evolves with classroom needs.