Child-AI Creativity Support Tools: A Co-Design Study with Children

Creative expression plays an essential role in the development of cognitive skills. Creativity Support Tools (CSTs) such as child-AI visual storytelling where children can verbally narrate and figuratively draw and express their stories are increasingly used to scaffold and support children’s creative expression, helping them overcome creativity decline or blocks such as the fourth-grade slump, a phenomenon where children’s creativity declines by grade 4. However, child-AI visual storytelling interfaces are not always designed to align with children's mental models and expectations, nor do they fully accommodate their developing cognitive and motor skills. For example, generative AI interfaces such as DALL-E have been used as examples of Creativity Support Tools for visual art and storytelling for children, however, they are not always designed to accommodate children’s needs and the diversity of their interactions in mind. The goal of this MS thesis is to understand children’s mental models of child-AI visual storytelling tools, especially with regard to interaction methods and the level of automation support needed from AI to support children’s creative storytelling expression. To answer our questions, we conducted four co-design sessions at the University of Washington’s KidsTeam with 7 children aged 8-13. The sessions were hybrid and lasted 90 minutes each. The main goal of the session was to understand children’s conceptual models related to how they might think of interacting with storytelling tools to express their ideas, and how they conceptualize distributing creating tasks between AI and themselves. We qualitatively analyzed data using thematic analysis to create a conceptual model of how children perceive creativity support tools and point out their preferences regarding the kind of interactions the interface can afford, main components of the model include: Negotiation, Division of Work, and Input Methods. Our work contributes a deeper understanding of children’s conceptual model of child-AI creativity support tools for storytelling in terms of automation, feedback, and interaction and will help inform the design of future child-AI creativity tools.