Building a Generative AI-based Agent to Cultivate Science Curiosity: An Intervention Study

• In Event: AERA Poster Session Friday 9:45am
  In Poster Session: STEM Education and Computational Learning Innovations

Abstract

Science curiosity is essential for student engagement and innovation but remains underdeveloped in many education systems. Traditional instruction often fails to meet the motivational and cognitive needs required to foster curiosity. With the rise of GenAI, researchers are exploring its potential to support curiosity-driven learning. This study designed a GenAI-powered educational agent based on reward-learning theory and evaluated its effects in a randomized controlled trial. Over three sessions, all participants read the same article; the control group (n = 70; 76% female) generated questions independently, while the experimental group (n = 67; 73% female) engaged in guided inquiry with the agent. Results suggest the GenAI intervention promoted both short-term and long-term curiosity, highlighting its potential as a scalable educational support.

Authors

• Zepeng Liu, Shanghai Jiao Tong University

• Jin Xie, Shanghai Jiao Tong University

• Kui Xie, University of Missouri

• Xin Tang, Shanghai Jiao Tong University