SkillTree: Scalable Personalized Learning with LLM-Linked Custom Knowledge Graphs

SkillTree: Scalable Personalized Learning with LLM-Linked Custom Knowledge Graphs is sponsored by IES - Institute of Education Sciences. Addresses the shortage of computer science teachers by developing an AI-powered platform that provides real-time evaluation and custom feedback to support personalized learning in high schools.

SkillTree: Scalable Personalized Learning with LLM-Linked Custom Knowledge Graphs | IES SkillTree: Scalable Personalized Learning with LLM-Linked Custom Knowledge Graphs From Seedlings to Scale Grants Program University of Illinois, Chicago (09/30/2025 - 09/29/2026) U.S. high schools are experiencing a computer science crisis.

As computing and AI literacy skills become increasingly important to be successful in the workforce, policymakers and educators in several states are responding by requiring at least one course in computer science (CS) in high school for graduation.

At the same time, there is a significant shortage of CS teachers nationwide, thereby restricting the opportunities students have for personalized instruction and attention and stretching existing teachers’ capacity (Shien, 2019; Werner & Chen, 2024a).

Modern automated technologies like generative AI and LLMs have demonstrated promise to support real-time evaluation and diagnosis of students’ knowledge and subsequently provide custom feedback and tutoring. Despite their promise, however, AI technologies for personalized learning are still largely un-trustable, owing to their ‘black-box’ operations and the inability to explain the rationale of outputs.

This project seeks to address this problem, opening the doors for highly scalable and trustable evaluation of students’ knowledge and subsequent delivery of formative feedback on performance. The proposed SkillTree will be an AI-based system that will deliver an automatically generated visual representation or ‘map’ of students’ knowledge, skills, and interest within a specific domain–specifically computer science for this project.

Leveraging advancements in industry AI research from the last few months with novel approaches—such as knowledge graphs, network databases, Chain-of-Thought (CoT), and Retrieval Augmented Generation (RAG)—this system will provide a personalized snapshot of a students’ real-time knowledge and experience.

Inspired by the design of skill trees commonly found in action-adventure and role-playing video games, as students perform new learning tasks their personal SkillTree will ‘light up’ or ‘level up’ new nodes in their network based on their demonstrated expertise. SkillTree will solve a perennial problem in education by supporting how teachers know each individual students’ prior knowledge, expertise, interests, and experiences.

With SkillTree’s assistance, a teacher can refine their teaching strategies. SkillTree will also contribute to persistent educational AI technology bottlenecks by forcing an LLM to use an independent underlying model of knowledge within a domain to draw its inferences. This underlying model can be readily validated by experts and educational standards in the field.

The research and development strategy will involve two activities: a comprehensive, audience Alignment Study and the development of a functional system prototype. These activities will result in three deliverables: a proof-of-concept report, a Phase Two R&D Plan, and an early-stage prototype. People and institutions involved Commissioner of Education Research Co-principal investigator Questions about this project?

To answer additional questions about this project or provide feedback, please contact the program officer. Questions about this project? To answer additional questions about this project or provide feedback, please contact the program officer.

Data Science for Education (DS4EDU) Instructional Strategies to Help Middle and High S... Condition of Education 2025 National Center for Education Statistics (NCES)