Proceedings of the International Group for the Psychology of Mathematics Education · Volume 1
Mario Lepore, Roberto Capone, Eleonora Faggiano
University of Salerno
⬇ Download PDFGenerative AI is increasingly entering mathematics classrooms, yet it may shift students’ work toward uncritical acceptance of generated outputs. To counter this risk, we analyse how learning by prompting can extend constructionist learning by making, supporting mathematical thinking within programming activities. We report a classroom‑based qualitative case study in an Italian scientific high school where small groups worked on a Python Turtle geometry task with a customised GenAI assistant. We trace how prompts evolved from vague requests to explicit mathematical specifications across prompt–debug–refine cycles. Results show that mismatches between intended and produced outputs triggered revisions that made constraints more explicit and supported decomposition of the geometric structure.
Computational thinkinggenerative AIconstructionismproblem based learning