Co-Ideation Across Time: Revitalizing Legacy Design Sketchnotes with Conversational AI Agents to Foster Intergenerational Collaboration

Paper Title

Co-Ideation Across Time: Revitalizing Legacy Design Sketchnotes with Conversational AI Agents to Foster Intergenerational Collaboration

Publication Info

• Topic area: AI-supported intergenerational knowledge sharing in design education.
• Keywords: AI Mentor, legacy sketchnotes, intergenerational collaboration, design ideation, conversational agents, LLMs, tacit knowledge, design education, AI-augmented Knowledge Objects, co-creation.

Background and Problem

• Problem / challenge: Legacy sketchnotes, which capture rich design rationales and inspirations, are underutilized due to their static nature, illegibility, and lack of contextual information, limiting their potential for knowledge exchange and ideation.
• Significance: Revitalizing legacy sketchnotes can bridge generational gaps in design education, enabling deeper understanding, creative ideation, and continuity within design communities.
• Motivation and related work: Prior work highlights the importance of process-focused design tools, example-based ideation systems, and AI-mediated mentorship. However, existing systems often fail to address the challenges of interpreting messy, handwritten artifacts or facilitating nuanced intergenerational collaboration.

Solution

• Proposed approach: Co-Ideation Across Time (CIAT), a system that integrates AI Mentors and AI-augmented Knowledge Objects to transform legacy sketchnotes into interactive tools for ideation and learning.
• Novelty:
  1. A pipeline for creating AI Mentors trained on human mentors’ publications, theses, and social media content.
  2. A method to convert legacy sketchnotes into AI-augmented Knowledge Objects for interactive exploration.
  3. Empirical evidence demonstrating how AI Mentors enhance comprehension, ideation, and intergenerational knowledge sharing.
• Procedure and key techniques:
  • Annotation Labeling Interface: Users upload sketchnotes, annotate regions, and generate AI Mentor responses.
  • AI Mentor creation: Combines academic publications, presentations, and social media content using Retrieval-Augmented Generation (RAG) and few-shot prompting.
  • AI-augmented Knowledge Objects: Annotated sketchnotes are stored as JSON files for interaction.
  • CIAT Interface: Node-based whiteboard interface for real-time brainstorming with AI Mentors and sketchnotes.

Results

• Concrete findings:
  • BLEU/ROUGE scores of AI Mentor responses were low (F1 = 0.2), but BERTScore was high (F1 = 0.8), indicating strong semantic alignment despite stylistic differences.
  • Participants shifted from mechanical to conceptual and metaphorical understanding of design concepts, e.g., Radical Atoms.
  • Self-reported creative engagement scores were high (median = 6/7), with variability in engagement tied to trust in the AI Mentor.
• Advantage over baselines:
  • Enhanced comprehension of spatial-temporal relationships in sketchnotes.
  • Reduced cognitive load and increased curiosity compared to static interpretation.
  • Nuanced ideation strategies enabled by domain-specific AI feedback.
• Experiments / evaluation:
  • Exploratory evaluation with 12 participants (8 male, 4 female, aged 18–34, with 4+ years of design experience).
  • Tasks included interpreting sketchnotes with/without AI Mentor and ideating using the CIAT interface.
  • Data collected via transcripts, surveys, interaction logs, and interviews.
• Limitations and future work:
  • Order effects in the study design may influence quantitative results.
  • Technical usability issues affected three participants.
  • Future work includes expanding to other domains, training multiple AI Mentors, and enabling group ideation.

Summary

This paper introduces CIAT, a system that leverages AI Mentors and AI-augmented Knowledge Objects to revitalize legacy sketchnotes for intergenerational collaboration. Empirical findings show that the system enhances comprehension, stimulates creative ideation, and fosters reflective learning across time. By bridging gaps between static design artifacts and interactive AI tools, CIAT contributes to design education and knowledge sharing. Future work will explore broader applications, multi-mentor perspectives, and collaborative ideation scenarios.