SparkTales: Facilitating Cross-Language Collaborative Storytelling through Coordinator-AI Collaboration

Paper Title

SparkTales: Facilitating Cross-Language Collaborative Storytelling through Coordinator-AI Collaboration

Publication Info

• Topic area: AI-assisted systems for cross-language collaborative storytelling in educational contexts.
• Keywords: Cross-language learning, collaborative storytelling, AI-assisted education, large language models, coordinator support, child engagement, bilingual education, multimodal materials, verbal engagement, human-AI collaboration.

Background and Problem

• Problem / challenge: Coordinators face significant cognitive and organizational burdens in cross-language collaborative storytelling, including managing language support, maintaining engagement, and addressing cultural differences. Existing tools lack systematic support for these tasks.
• Significance: Enhancing cross-language collaborative storytelling can improve children's language learning, cultural understanding, and expressive abilities, while reducing coordinators' workload.
• Motivation and related work: Prior studies have explored storytelling and AI-assisted education but lack focus on cross-language collaborative storytelling, particularly in online contexts. Existing systems often fail to balance personalization, cultural sensitivity, and engagement, leaving gaps in coordinator support.

Solution

• Proposed approach: SparkTales, an LLM-based intelligent assistant designed to support coordinators in cross-language collaborative storytelling by generating story frameworks, questions, and multimodal materials while tracking engagement and providing feedback.
• Novelty:
  1. First AI system specifically designed for coordinators in collaborative storytelling.
  2. Integration of individual and common characteristics to balance personalization and shared context.
  3. Dynamic generation of story frameworks, questions, and comprehension materials tailored to children’s profiles.
  4. Automated engagement reviews and feedback to optimize future sessions.
• Procedure and key techniques:
  • Configuration Module: Allows coordinators to set target vocabulary and children’s features.
  • Individual and Common Characteristic Summarization Modules: Extract and summarize personalized and shared traits using semantic matching and reasoning.
  • Collaborative Storytelling Support Module: Generates bilingual story frameworks, guiding questions, and multimodal materials.
  • Review and Feedback Module: Tracks children’s participation and provides actionable insights for refinement.

Results

• Concrete findings:
  • Coordinators rated SparkTales highly for functionality (4.53/5), performance (4.47/5), and usability (4.39/5).
  • Children demonstrated high verbal engagement: average productivity (20.87 words per response), lexical diversity (17.07 unique words), and topical relevance (1.74/2).
• Advantage over baselines:
  • Reduced coordinator workload by automating material preparation, question generation, and engagement tracking.
  • Enhanced children’s participation by tailoring content to individual and shared characteristics.
• Experiments / evaluation:
  • Participants: 8 teachers and 16 children aged 7–11, paired for bilingual storytelling (Chinese-English).
  • Metrics: Technology Acceptance Model (TAM) for coordinators; six dimensions of verbal engagement for children.
  • Procedure: Pre-activity configuration, storytelling tasks, and post-activity reviews.
• Limitations and future work:
  • Limited adaptability to diverse age groups, languages, and larger participant pools.
  • Challenges in capturing fine-grained individual and shared traits.
  • Short-term evaluation lacks insights into long-term impacts.

Summary

SparkTales is an innovative AI-assisted system that supports coordinators in cross-language collaborative storytelling by automating key tasks and enhancing children’s engagement. Evaluation results demonstrate its effectiveness in reducing workload and fostering active participation, particularly through personalized and shared content generation. Future research should explore broader applicability across diverse demographics, languages, and long-term use scenarios.