Diana Gojcaj
Research Mentor: Xuechen Liu
Mentor Department: Educational Studies, Education
Author(s): Diana Gojcaj, Keisha Karla
Session: Session 2 (10:00 AM – 10:50 AM)
Presentation Type: Poster 36
Abstract
Dialogical reading, in which children are actively engaged in conversation during shared book reading, is a powerful approach for supporting language development and content understanding. Recent advancement of Generative AI presents a promising potential for engaging in and scaffolding book reading experiences with children by posing questions, responding to children’s ideas, and sustaining dialogue around story content. While prior studies have demonstrated the potential of AI-supported reading, there is limited research examining the interactional process itself—how children engage with AI characters during reading and the quality of those interactions. Drawing on data from a larger research project that developed an AI character embedded in an interactive science eBook, this exploratory study examines how children interact with an AI partner and the quality of the conversation during shared reading. We will analyze transcripts from twelve child–AI reading sessions involving children in first and second grade, in which the AI character prompted discussion through open-ended, content-related questions as children progressed through the eBook. More specifically, we will take a both qualitative and quantitative analysis approach to examine 1) the types of questions AI asks; 2) children’s responses to its questions; 3) the accuracy of children’s responses to the AI’s prompts and 4) the effectiveness of AI’s scaffolding. Our findings will characterize patterns in children’s engagement during AI-supported dialogic reading interactions at this early stage of research. Findings characterize how children express and elaborate their ideas, engage with scientific concepts, and participate in dialogue with an AI character during eBook reading, as well as how the AI’s scaffolding supports or constrains these interactions. These insights contribute to a more process-oriented understanding of child–AI interaction and offer design implications for future AI systems intended to support interactive and dialogic reading experiences.
