SBIR Phase II: An online learning and assessment platform for sophisticated and secure exams

This project develops an artificial intelligence (AI) powered online platform that enables instructors to create sophisticated, automatically graded, and individualized test questions. The platform supports improved student learning through repeated practice with immediate feedback, strengthened exam security through question randomization and restricted access to outside resources. This Small Business Innovation Research (SBIR) Phase II project addresses the challenge of using artificial intelligence (AI) to generate reliable educational assessments at scale. Large language models can automate content creation but frequently produce plausible yet incorrect outputs, making them unsuitable for direct use in high-stakes testing. Building upon a successful Phase I proof of concept, this project employs a novel two-stage approach: first, an AI model converts an instructor's plain-language description into validated computer code that encodes both question generation and grading logic; second, this verified code produces unique, randomized question instances for each student. Because the AI generates reusable code rather than directly grading student work, and because the code undergoes rigorous automated and human validation before use, the approach eliminates reliability concerns associated with direct AI assessment. The Phase II research objectives are to expand the AI engine to generate varied question types including graphical, programming, and data-analysis formats; enable question creation from uploaded course materials; implement an automated validation step in which a second AI model solves generated questions to verify correctness; develop a visual editing interface for non-technical users; and conduct iterative user studies measuring usability and pedagogical effectiveness. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.