Empowering Children’s Dialogic Reading with a Bilingual LLM-Powered Conversational Agent: A Design Probe Study

• In Event: AERA Roundtable Session Friday 11:45 am - Gold 4
  In Roundtable Session: Self-Regulation and Emotional Development in the age of AI

Abstract

Dialogic reading has been widely recognized for enhancing children's language and literacy skills, yet it is difficult in English-as-a-foreign-language (EFL) homes with limited adult proficiency. This study presents Storio, an interactive e-book with Mia, a bilingual large-language-model conversational agent that emulates dialogic scaffolds. Used as a design probe, Storio engaged EFL children (N = 17) and gathered feedback from parents (N = 19) and educators (N = 2). Findings indicate that Mia effectively supported language production, encouraged interactive reading, and promoted sustained engagement and language development. The study highlights the potential of LLM-driven agents in supporting multilingual learners and offers design insights for developing inclusive and adaptive agent-mediated reading tools.

Authors

• Feiwen Xiao, University of Pennsylvania

• Jiaju Lin, Pennsylvania State University

• Zhaohui Li, Pennsylvania State University

• Xiaohan Zou, Pennsylvania State University

• Chih-Yi Chen, Pennsylvania State University

• Dandan Yang, University of California - Los Angeles

• Wenting Zou, Pennsylvania State University

• Jinjun Xiong, University at Buffalo - SUNY