2026 ASEE Annual Conference & Exposition

Knowledge graphs have traditionally been used to effectively visualize, organize and structure course content by illustrating which concepts support and connect to one another. This work explores a web app extension of knowledge graphs, called Knowledge Grapht, aided by generative AI, and designed with future causal inference in mind to support instructors and students alike. This project enables students to utilize structured class knowledge graphs created by their instructors where each node reflects a concept and each edge encodes the pre-requisite structure. Students can take quizzes for each topic with a mix of instructor curated questions and AI generated questions with each topic's proficiency level dynamically adjusted according to quiz performance and review frequency. Students are encouraged to review topics they struggle with more frequently through a spaced-repetition schedule that emphasizes long-term retention, inspired by validated techniques like the Leitner Box. Instructors also benefit from using Knowledge Grapht as this quiz data is used to calculate and visualize population-level class statistics of concept mastery, as well as individual student proficiencies to help them navigate topics they may be struggling in. This project evaluated the integration of Knowledge Grapht in CMSI courses to examine relationships between topic-level mastery and final exam performance. The platform was deployed during regular course activities, and usage data from both students and instructors were analyzed alongside performance on graded assessments. The analysis examined correlations between practice activity, concept-level proficiency, and overall course achievement. Results revealed a positive correlation between students’ concept-level mastery scores and final exam performance, with higher engagement in targeted practice and spaced review associated with stronger overall course outcomes. Findings from this implementation provided insight into how AI-assisted knowledge graph systems can support personalized learning and inform future approaches to course design.