Adaptive AI Tutor Improves Certain Students’ Learning Outcomes in High School Physics (Poster 7)

• In Event: AERA Poster Session Friday 1:30 p.m.
  In Poster Session: Advanced Technologies for Learning Poster Session 2

Abstract

This paper describes a randomized controlled study that designed and tested an AI-powered adaptive tutoring system for high school physics instruction, specifically in developing regions. We gathered test scores and non-cognitive information to investigate the impact of AI-generated learner instruction on academic achievement and psychological-behavioral propensities. The trial involved 212 first-year high school students divided into three groups receiving personalized AI feedback, control, and another treatment. Treatment with AI-generated content significantly improved academic performance, particularly for students with lower pre-test scores and higher levels of self-regulation or grit. The result proves that the AI tutoring system designed by the study effectively enhances high school physics learning outcomes and can promote educational equity without investing in expensive electronic devices.

Authors

• Xusheng Dai, Tsinghua University

• Zhaochun Wen, Tsinghua University

• Jianxiao Jiang, Tsinghua University

• Yu Zhang, Tsinghua University