Illuminate: An Educational AI Science Sandbox​

Christian Rasmussen

Co-Presenters: Brodie Berger, Travis Matos, Ibrahim Khan Shovo

College: Hennings College of Science Mathematics and Technology

Major: MS.COMPUTER/SCIENCE

Faculty Research Mentor: Kwak, Daehan  

Abstract:

Title: Illuminate: An Educational AI Science SandboxAuthors: Brodie Berger, Travis Matos, Christian Rasmussen, & Ibrahim Khan ShovoAbstract:Artificial Intelligence (AI) has become a pivotal tool in education; however, current implementations often impede deep learning by providing direct answers rather than fostering critical thinking. Illuminate aims to solve this problem. Illuminate is an interactive, open-ended learning environment designed to teach core concepts in Physics, Chemistry, and Electricity. Unlike traditional educational software that relies on scripted lessons, Illuminate utilizes a sandbox architecture that encourages users to discover scientific principles through direct manipulation and experimentation.This application will feature two different modes, such as a gamified mode with structured levels/lessons, as well as the sandbox mode that allows for free experimentation. The core of this project will be the integration of an AI companion known as “Lumi”, powered by a custom hosted small language model. This agent will observe the players' actions in real time and be able to not just answer direct questions but provide hints and guidance to promote learning without giving direct solutions. The system is built with Matter.js for 2D physics simulations, a logic-based solver for electrical circuit analysis, and a hybrid SQL-AI backend that dynamically verifies and renders chemical compound properties. By combining open-ended experimentation with AI-driven guidance, Illuminate provides a framework for educational software that fosters curiosity and active learning rather than simply delivering answersKeywords: Educational Technology, Science, Artificial Intelligence, Interactive Simulation, Small Language Models