FURI | Spring 2025

Adaptive Learning for Computer Science Education

Mastery learning empowers students to progress at their own pace, ensuring they achieve deep understanding before advancing. This work-in-progress paper introduces an adaptive learning system for introductory computer science courses that leverages the Bayesian Knowledge Tracing (BKT) algorithm to model and support individual student learning trajectories.

Unlike traditional models that use fixed quizzes or uniform assessments, this system continuously estimates a student’s knowledge state over time using BKT. It tracks key indicators — such as time spent on tasks, response correctness, and attempt history — to dynamically adjust content difficulty, suggest targeted review material, and provide real-time feedback. Rather than merely accommodating different learning styles, the system emphasizes data-driven personalization and automated mastery pathways.

Course content is modular, with mastery assessed per module through adaptive problems tailored to a student’s current estimated knowledge. Question parameters — including format (e.g., multiple choice, coding), cognitive level (recall to analysis), and distractor effectiveness — are used to fine-tune both content sequencing and predictive accuracy. Through ongoing interaction, the system identifies conceptual gaps early, reinforces weak areas via spaced repetition, and prevents knowledge decay.

By integrating BKT with continuous feedback loops and learning analytics, the system offers a scalable solution for mastery-based learning. It aims to improve long-term retention, foster deeper understanding, and ensure that every student progresses only when mastery is demonstrably achieved.