News Roundup: AI Tutor Helps Address Math Learning Gaps

Researcher Neil Heffernan has been awarded a three-year grant from the U.S. Department of Education’s (DOE) Institute of Education Sciences to develop an artificial intelligence tutor that can help middle-school students learn math while doing homework. Heffernan is the William B. Smith Professor of Computer Science and director of the Learning Sciences and Technologies Program.

Under a cost-sharing agreement, the DOE’s Institute of Education Sciences will contribute $3,749,600, or 90.89 percent of the funding, to the $4.1 million project, and WPI will contribute $375,815, or the remaining 9.11 percent. The project will integrate the AI tutor into ASSISTments, a free digital learning platform that has been used by more than one million students across the country and was developed by Heffernan and his wife, Cristina Heffernan, a former math teacher. The project aims to address the needs of students who have fallen behind in class but cannot afford private tutoring services.

“Tutors are very effective at helping students learn math and succeed in class, but the cost of private tutoring services is beyond students from low-income backgrounds,” Neil Heffernan says. “This leads to a persistent learning gap between lower-income students and students from families that can afford tutoring. A free AI tutor that students could access after school while doing homework would help address this gap and enable lower-income students who have fallen behind in class to catch up to their peers, be more engaged with their lessons, and succeed as they learn the concepts needed to advance to higher-level math.”

Heffernan will lead a team of psychology and learning science researchers, teachers, and education experts to develop CAIT (pronounced “Kate”), a Conversational AI Tutor that they will integrate into ASSISTments and then test in classroom settings. ASSISTments previously has been shown to positively impact students’ math achievement and educational equity in an independent evaluation of nearly 6,000 middle-school students in North Carolina. In addition to Heffernan, WPI researchers involved in the project will include Assistant Professor Stacy Shaw of the Department of Social Science and Policy Studies, Computer Science Research Scientist Li Cheng, and Assistant Professor Adam Sales of the Department of Mathematical Sciences. The WPI team will work with the ASSISTments Foundation, an independent nonprofit organization founded by the Heffernans to manage and market the platform. WestEd, an education research organization, will also partner to research and develop the use of CAIT in educational settings.

The project is possible because of recent advances in generative AI, a type of artificial intelligence that uses complex algorithms and is trained to find patterns in data in order to create new material. CAIT will have a “conversational interface” powered by natural language processing, a field of AI that enables machines to understand and generate human language. As a result, students will be able to speak or write questions for CAIT, and CAIT will respond with personalized replies. By monitoring students’ work, CAIT will be able to identify challenges each student is encountering and provide encouragement and additional problems for students to solve.