Author Information
Ting Xu, The Hong Kong University of Science and Technology, ChinaXian Xu, Lingnan University, Hong Kong
Abstract
The rapid adoption of generative artificial intelligence (AI) in education does not automatically enhance children's creativity. Prevailing prompt–response interaction patterns often position children as passive consumers, risking the homogenization of their distinctive imagination and narrative style. This study proposes and examines child-guided co-narration, an alternative paradigm that repositions the child as narrative authority and the AI as a responsive partner to children's unstructured, play-based expressions, such as spontaneous drawing, block building, and improvised telling. We instantiated the paradigm in StorySketcher, a multi-agent AI system that transforms children's sketches and voices into animated, multilingual stories while preserving creative agency through Socratic prompting. A two-phase mixed-methods study combined participatory design (n = 12 children, aged 5–8) with a four-week quasi-experimental educational study (n = 36 children, aged 6–8) comparing the child-guided system with a prompt–response AI storytelling application. Narrative agency was operationalized as the child-initiated decision ratio; narrative complexity through story-grammar and cohesion coding; originality through consensual expert assessment; and dynamic visual literacy through a purpose-built rubric. Results show that the child-guided condition produced significantly higher narrative agency, more complex and more original stories, and stronger dynamic visual literacy. We argue that treating the child as guide reframes AI from ghostwriter to an extension and mirror of children's imagination, offering a framework for protecting and enhancing children's artistic and narrative expression in the AI era.
Comments
Powered by WP LinkPress