Supporting Future-Ready Math Classrooms with ProductiveMath: A Web-based, Human-in-the-Loop Generative AI Tool

• In Event: AERA Roundtable Session Wednesday 1:45 pm - Gold 3
  In Roundtable Session: Integrating Artificial Intelligence in Teaching and Learning

Abstract

Productive Failure (PF) promotes deep learning and critical thinking by engaging students in problem-solving before instruction. However, designing PF-suitable problems is time-consuming for teachers. To support them, we developed ProductiveMath, a generative-AI tool, through four iterative studies: (1) creating a PF-quality rubric to prompt GPT-4o; (2) validating 60 AI-generated problems; (3) evaluating ProductiveMath’s design with teachers; and (4) implementing it in five schools with 66 students. Students reported increased engagement, collaboration, and problem-solving, though some faced challenges with reading and complexity. All teachers expressed interest in continued use of the tool, highlighting its ease and alignment with PF. ProductiveMath lightens teachers’ design burden while maintaining their control. This study shows how human-AI collaboration can advance future-ready learning in today’s classrooms.

Authors

• Seyedahmad Rahimi, University of Florida

• Maryam Babaee, University of Florida

• Salah Esmaeiligoujar, University of Florida

• Deniz Ercan, University of Florida

• Ran Gao, University of Florida