Artificial Un-intelligence? How Mathematics Coaches Utilize Inaccuracy in an AI Feedback Tool

• In Event: AERA Roundtable Session Friday 3:45 pm - Gold 2
  In Roundtable Session: Supporting Mathematics Teachers: Leadership and Retention

Abstract

How coaches provide feedback and elicit teacher reasoning affects relational trust. As Artificial Intelligence enables new forms of scalable instructional feedback, questions remain about accuracy. This analysis, part of a larger study on how mathematics instructional coaches utilize automated feedback, focuses on how coaches responded when an AI tool missed or falsely identified elements in recorded lessons. In several cases, coaches used these AI errors as pedagogical opportunities to elicit teacher reasoning or maintain a nonjudgemental stance. While AI continues to improve in educational tasks, it will likely still make mistakes. These findings have implications for how coaches and other educators might leverage those errors in service of professional learning.

Author

• Jim Malamut, Stanford University