Math Learning and Teaching with Generative AI
Use rational number learning to evaluate AI-generated feedback, remediate misconceptions, and identify the human factors driving student success.
Project Leads
Jing Tian, Assistant Professor, Department of Psychology
Juntao Chen, Assistant Professor, Department of Computer and Information Sciences
Peter Zangari, AI/Finance, Senior Business Leader and Instructor at the Gabelli School of Business
Project Description
While AI-supported math instruction is not new, prior generations of educational technologies have seen limited adoption despite demonstrated benefits. Generative AI marks a qualitative shift: it is highly accessible, adaptive, and already embedded in students’ everyday learning practices. Emerging evidence suggests that generative AI can enhance learning through immediate, personalized feedback and practice problems to address misconceptions. At the same time, there are significant concerns. Students may over-rely on AI-generated answers, engage less deeply with materials, or fail to evaluate the correctness of outputs. Teachers also report mixed attitudes toward integrating AI into teaching and learning, ranging from enthusiasm to concerns about accuracy, pedagogy, and ethics.
To prepare students and educators for this new landscape, foundational research is needed to understand how cognitive and motivational processes unfold when learning occurs through interactions with the AI systems. The proposed project serves as an initial step of this inquiry by focusing on rational number learning (fractions, decimals, and percentages) as a testbed. Mastery of rational numbers is a well-documented challenge and a key bottleneck for success in more advanced mathematics
This project pursues three questions: (1) To what extent is AI-generated feedback conceptually comparable to feedback from human instructors? (2) Can AI-generated practice effectively diagnose and remediate student misconceptions? (3) Which student-level factors (e.g., prior knowledge, motivation) shape learning outcomes in AI-supported environments? We position this work as a series of pilot studies to establish feasibility, refine measures, and generate effect size estimates for future large-scale research.
The project team integrates complementary expertise: a developmental psychologist specializing in cognitive and motivational processes in math learning, an AI expert in generative systems, and a practitioner experienced in applying AI in instructional contexts. This collaboration enables a rigorous and ecologically valid examination of human–AI interactions in learning.