Artificial intelligence for humanizing and enhancing the learning of proofs

This National Science Foundation (NSF) STEM Education (CFDA 47.076) Project Grant award provides $749,996 to the University of Florida to develop an innovative artificial intelligence model that can provide feedback on student proofs in proof-oriented mathematics courses. The project aims to improve the proof-writing abilities of undergraduate students by training the AI model to give immediate, iterative, research-based feedback on student proofs.

The project involves sub-awards to St. Olaf College, Education Development Center, Inc., and Marquette University, who will contribute expertise in mathematics, mathematics education, and project management to support the overall research and development activities. Key deliverables include the creation of a website that allows students to write proofs and receive AI-generated feedback, as well as the development of an open-source "constitutional AI" approach to align the AI model with educational values and priorities. The project will advance understanding of how students learn to write proofs and how AI can be leveraged to support STEM education.

Abstract

This project aims to serve the national interest by improving the proof-writing abilities of undergraduate students. This level II iuse project, part of the engaged student learning track, intends to accomplish this by training an innovative artificial intelligence model to provide feedback on student proofs to support both college students in proof-oriented mathematics courses and the post-secondary faculty who are teaching them. The project will create a website to demonstrate the model: students will be able to use the website to write proofs, receive immediate feedback on their proofs, and revise and resubmit new drafts for feedback. This model is not intended to replace the work of mathematicians as teachers, but to offer an important tool to faculty - one that gives students immediate, iterative, researched-based feedback on their written proofs in ways that exceed what is practical for professors to provide, and also attends to the needs of diverse learners. To reach its goals, the project will take a constitutional AI approach, convening a group of experts in proofs, equity, and college-level mathematics teaching and learning to build a constitution that will guide the development and training of a large language model (LLM) specifically trained to give feedback on student proofs that is in line with the constitution. Project research will focus on how the LLM serves students and faculty. The constitution will drive the development of the LLM and will be available for use to support future AI interventions and educational technology work. This work will advance understanding of how students learn to write proofs, how artificial intelligence can be used to support student learning, and how constitutional AI approaches can be used to align artificial intelligence with educational values and priorities. The NSF iuse: EDU program supports research and development projects to improve the effectiveness of stem education for all students. Through the engaged student learning track, the program supports the creation, exploration, and implementation of promising practices and tools. 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.- subawards are planned for this award.