Textbook, Teacher, and AI-Generated Word Problems and Third Graders’ Evaluations of Them

• In Event: AERA Roundtable Session Wednesday 3:45 pm - Gold 2
  In Roundtable Session: Curricular Resources for Mathematics: Friends or Foes?

Abstract

We explored AI prompts for creating word problems for third graders that would be relevant and similar to textbook and teacher-generated word problems. Comparisons between AI-generated, teacher-generated, and textbook word problems revealed that the problem lengths were similar, and all included contexts related to collections. The AI problems used different keyword phrases than the teacher and textbook problems. We had 8 students determine which problems (presented in five pairs) were written by AI or their teacher. Their reasoning revealed assumptions about how AI and teachers generate problems, such as thinking AI would use unrealistic contexts or strange words. While AI can create problems that many students cannot distinguish from traditional sources, careful attention to prompting and language is crucial.

Authors

• Laura Bofferding, Purdue University

• Yi Zhu, Purdue University

• Sezai Kocabas, Purdue University