Towards Understanding Children’s Collaborative Interaction Patterns in Child-AI Co-creative Interfaces

Paper Title

Towards Understanding Children’s Collaborative Interaction Patterns in Child-AI Co-creative Interfaces

Publication Info

• Topic area: Child-AI interaction in co-creative storytelling interfaces
• Keywords: Child-AI collaboration, co-creativity, participatory design, generative AI, storytelling, child-centered design, interaction patterns, creativity support tools, collaboration profiles, child development

Background and Problem

• Problem / challenge: Existing AI co-creativity tools often fail to align with children’s developmental needs and natural collaboration behaviors, risking a reduction in meaningful creative opportunities.
• Significance: Understanding how children naturally collaborate with AI can inform the design of interfaces that scaffold creativity while preserving children’s agency and self-efficacy.
• Motivation and related work: Prior studies have explored AI’s role in supporting creativity but have not deeply examined children’s natural collaboration patterns with AI. Research has called for deeper exploration of child-AI collaboration, particularly in co-creative contexts like visual storytelling.

Solution

• Proposed approach: The Child-Centered Co-Creative AI (CCAI, “Kai”) framework, informed by participatory design sessions with children, to guide the design of child-AI co-creative interfaces.
• Novelty:
  1. Introduction of four child–AI collaboration profiles: Independent, Child-Driver + AI-Refiner, AI as Inspirer, and Child-Initiator + AI-Transformer.
  2. Identification of seven types of AI contributions and children’s preferences for each.
  3. Development of the CCAI framework with four design dimensions: child-centered multimodal input methods, metacognitive scaffolds for AI suggestions, embodied affordances for co-creative AI, and reducing cognitive load during co-creation.
• Procedure and key techniques:
  • Conducted four 90-minute participatory design sessions with seven children (ages 8–13) using Cooperative Inquiry methods.
  • Analyzed children’s interaction patterns with an AI drawing tool, focusing on technical and conceptual contributions.
  • Used affinity diagramming to identify collaboration profiles, interaction strategies, and alignment challenges.

Results

• Concrete findings:
  • Four collaboration profiles were identified, reflecting dynamic patterns of child-AI interaction.
  • Children preferred AI contributions that refined or extended their ideas without overriding their original intent.
  • Seven types of AI contributions were mapped to children’s preferences, including detail refinement, shading, and content expansion.
• Advantage over baselines:
  • Provides a nuanced understanding of children’s preferences for AI contributions, addressing gaps in prior research that treated AI as a passive generator rather than an active collaborator.
• Experiments / evaluation:
  • Conducted with an intergenerational co-design group using a participatory design approach.
  • Evaluated children’s interactions with the Pikaso Freep!ck AI drawing tool across two sessions focusing on technical and conceptual help.
• Limitations and future work:
  • Small sample size (n=7) limits statistical generalizability; findings are intended as foundational insights.
  • Future work should test the framework with diverse populations and extend it to other aspects of child-AI collaboration, such as communication strategies and task negotiation.

Summary

This study investigates how children naturally collaborate with AI in co-creative storytelling contexts. Through participatory design sessions, the authors identified four collaboration profiles and seven types of AI contributions, revealing children’s preferences for refinement over transformation. The Child-Centered Co-Creative AI (CCAI) framework was developed to guide the design of future child-AI interfaces, emphasizing multimodal input methods, metacognitive scaffolds, embodied affordances, and reduced cognitive load. These findings provide actionable insights for designing AI systems that support children’s creativity while respecting their agency.