MuseForge: Enhancing Creative Learning in Digital Museum Education with Generative AI

Paper Title

MuseForge: Enhancing Creative Learning in Digital Museum Education with Generative AI

Publication Info

• Topic area: Creative learning in digital museum education using Generative AI.
• Keywords: Creative learning, digital museum, Generative AI, non-formal education, Resnick’s five-stage model, creative self-efficacy, engagement, motivation, cultural knowledge, human-AI interaction.

Background and Problem

• Problem / challenge: While creative learning is well-integrated into on-site museum education, its structured application in digital museum contexts remains underexplored. Existing digital museum platforms often lack scaffolding for sustained creative engagement, and most Generative AI (GenAI) research focuses on isolated creative stages or formal educational settings.
• Significance: Bridging this gap can enhance engagement, self-expression, and learning outcomes in non-formal educational environments, such as digital museums, which are increasingly relevant in expanding access to cultural content.
• Motivation and related work: Prior research has demonstrated the potential of GenAI in supporting creativity, but it has largely been confined to formal education or specific creative stages (e.g., ideation or production). Non-formal settings like museums require systems that support open-ended, self-directed, and iterative creative processes across all stages of learning.

Solution

• Proposed approach: MuseForge, a GenAI-powered platform, integrates Resnick’s five-stage creative learning model (Imagine, Create, Play, Share, Reflect) to scaffold personalized and dynamic creative learning experiences in digital museum education.
• Novelty:
  1. Development of MuseForge, a GenAI-supported system grounded in a five-stage creative learning framework.
  2. Comprehensive evaluation through a between-subjects study comparing MuseForge to a Baseline system.
  3. Insights and design implications for future creative learning systems in non-formal educational settings.
• Procedure and key techniques:
  1. Conducted a formative study with a prototype system to identify user needs and challenges.
  2. Designed MuseForge with features like thematic prompts, context-aware AI support, and reflective tools.
  3. Integrated a Retrieval-Augmented Generation (RAG) pipeline for domain-specific grounding of AI outputs.
  4. Evaluated MuseForge in a between-subjects study with 32 participants, comparing it to a Baseline system.

Results

• Concrete findings:
  • MuseForge significantly improved learning motivation (Mean = 5.44 vs. 4.20, p = 0.007, Effect Size = 0.47), engagement (Mean = 6.12 vs. 4.58, p < 0.001, Effect Size = 0.72), and creative self-efficacy (Mean = 4.81 vs. 2.06, p = 0.003, Effect Size = 0.53).
  • Knowledge acquisition showed no significant difference between MuseForge and the Baseline system (p = 0.351).
  • Participants expressed a higher willingness to reuse MuseForge (p < 0.001), though ease of use and ease of learning were not significantly improved.
• Advantage over baselines: MuseForge outperformed the Baseline system in fostering motivation, engagement, and creative self-efficacy, while maintaining comparable knowledge acquisition.
• Experiments / evaluation:
  • Participants (N=32) were divided into two groups: one using MuseForge and the other a Baseline system.
  • Metrics included learning motivation, engagement, knowledge acquisition, creative self-efficacy, and usability, assessed through pre/post-tests, surveys, and interviews.
  • Statistical analysis (Mann-Whitney U test) confirmed significant improvements in key areas for MuseForge users.
• Limitations and future work:
  • Limited to a single thematic domain (maritime art) and a relatively small, homogeneous participant group.
  • Desktop-based interface lacked natural sketching affordances.
  • Future work should expand to diverse cultural contexts, improve onboarding, and enhance AI grounding with larger knowledge bases.

Summary

MuseForge integrates Generative AI with Resnick’s five-stage creative learning model to enhance creative learning in digital museum education. A between-subjects study demonstrated significant improvements in learning motivation, engagement, and creative self-efficacy compared to a Baseline system, while maintaining comparable knowledge acquisition. The system’s design emphasizes structured scaffolding, learner agency, and cultural grounding, offering insights for future creative learning platforms in non-formal educational settings. Future work will address scalability, diverse content areas, and improved usability for broader audiences.