Learning Through Questioning in Child-AI Interactions: A Comparative Study of AI-Powered Agents Embodying Diverse Epistemic and Functional Attributes

• In Event: Child Development and Interactions

Abstract

This study examines preschoolers' questioning behaviors when interacting with two AI agents: an intelligent voice speaker and a social robot. 128 six-year-olds participated in a "You Ask and I Answer" task. We chose a social robot and an intelligent voice speaker as test cases due to their widespread use and divergent epistemic and functional attributes. Results show children asked more information-seeking questions to the voice speaker but more overall and functional questions to the robot. Moreover, children with richer prior experience with social robots asked more questions and more information-seeking questions to these agents. These findings highlight children's sensitivity to AI agents' attributes and inform the development of AI-powered educational tools.

Authors

• Suyang Wang, Capital Normal University

• Na Zhang, Capital Normal University

• Chengcheng Wang, Capital Normal University

• Manni Ma, Capital Normal University

• Xianan Lu, Capital Normal University