Xuechen Liu, Yuqing Xing, Youran Chen, Wenfei Pei, Jinshil Choi, and Ying Xu
Our goal is to investigate the influence of large language models (LLMs) and generative AI on children's education, particularly in the context of story-based learning. As a first step, we've launched a research study where children co-create STEM-oriented narratives with either a conversational AI agent or a human partner.
Children’s Narrative Participation in Story Co-Creation with AI vs. Human Partners: Response Styles and Narrative Features. https://doi.org/10.1145/3713043.3731486
Figure: Illustration from Xu, Y. Large Langauge Models to Support Learning Through Storytelling. Retrieved from https://ying-xu.com/research/ai-math/
Our goal is to investigate the influence of large language models (LLMs) and generative AI on children's education, particularly in the context of story-based learning. As a first step, we've launched a research study where children co-create STEM-oriented narratives with either a conversational AI agent or a human partner.
We aim to understand how children respond, reason, and imagine within these interactions. Our findings show that while children generally communicate directly with both AI and human storytellers, they tend to provide briefer, less elaborated responses with AI, and more extended, imaginative contributions with humans—especially among older children. These insights underscore the importance of tailoring AI design to children’s developmental and narrative needs.
To make AI storytelling more responsive and educationally beneficial, we’re exploring approaches such as theory-driven prompt engineering, retrieval-augmented generation (RAG), and fine-tuning. These techniques help safeguard the AI-generated dialogue, ensuring it aligns with pedagogical goals and fosters narrative competence, creativity, and STEM thinking.
This research not only sheds light on how young children engage with generative AI but also offers design principles for creating developmentally attuned, culturally sensitive, and curiosity-driven AI learning companions.