Toward Personalizable AI Node Graph Creative Writing Support: Insights on Preferences for Generative AI Features and Information Presentation Across Story Writing Processes

Research Background and Issues

• Identified Problems or Challenges: The authors recognize that one of the key purposes of storytelling is to convey moral implications. However, current generative AI tools primarily focus on technical support (e.g., language generation) or creative writing, neglecting the necessity of helping authors reflect on the moral implications conveyed in their stories. Additionally, existing tools lack personalization, cultural context understanding, and effective information representation.
• Significance: Stories can enhance social empathy and shape societal values by conveying moral lessons. Tools that support this aspect of storytelling can help authors express moral implications more clearly and adapt to readers from diverse cultural backgrounds.
• Research Motivation and Related Work: While previous studies have explored the potential of generative AI in creative writing, they have failed to integrate graphical visualization with generative AI to meet authors' dynamic needs throughout the creative process. Particularly in scenarios involving moral implications, tool support remains insufficient.

Solution

• Proposed Method or Solution: The authors designed and implemented a FigJam-based plugin called StoryNode, which integrates event node graph editing functionality, generative AI (such as LLM-driven character or audience simulation), and image and audio generation.
• Innovations:
  • Provides a unified interface allowing authors to freely switch between text and graphical formats.
  • Enhances personalization features, such as customized node and link styles to meet different authors' visualization needs.
  • Combines LLM-driven character and audience simulation to generate writing suggestions compatible with diverse cultural contexts.
  • Offers broader functionality—from logical visualization and creative generation to technical support—adapting to the dynamic needs of the creative process.
• Implementation Steps and Key Technologies:
  1. Embedding the StoryNode plugin into an online whiteboard tool (FigJam) to provide customizable editing functionality for node graphs.
  2. Integrating generative AI tools GPT-4o, Dall-E 3, and Suno v3 into the plugin for text, image, and audio generation.
  3. Enabling plugin-supported features, including converting continuous text into node graphs, visualization customization, and drag-and-drop node linking operations on the whiteboard.
  4. Iterative design based on data collected during user research and external evaluations.

Research Outcomes

• Specific Results:
  • The StoryNode plugin developed by the authors was validated through user research (N=14) and external evaluations (N=19) for its ability to support the conveyance of moral implications and improve the overall creative experience.
  • Achieved diverse visualizations of story logic and information relationships.
  • Found that combining event node graphs with generative AI is complementary in the creative process, supporting both logical reflection and inspiring creative ideas.
• Advantages:
  • Increased productivity: Integrated functionalities reduce the cost of switching between multiple tools.
  • Enhanced personalization support: The customization capabilities of node graphs significantly improve the flexibility of information presentation.
  • Ensured quality: External evaluations indicate that tasks completed using StoryNode are better at conveying story implications and exhibit clearer logic compared to traditional tools.
• Experimental or Evaluation Results:
  • In user research, StoryNode outperformed Twine significantly in the "Creativity Support Index" (CSI) scores, particularly in "expressiveness" and "exploration."
  • External evaluations showed that works produced using StoryNode were more recognized for conveying moral implications and logical clarity, with no significant cultural bias observed.
• Limitations and Future Directions:
  • Limited scope: The study mainly focuses on story outlining and shorter non-branching narratives.
  • Writing task constraints: It does not cover the complete book creation cycle, such as evaluating tool capabilities from a long-term perspective.
  • Future research directions include supporting more complex branching narrative designs, expanding cultural background samples, and addressing usability barriers caused by overly complex tool interfaces.

Through this study, the authors provide important goals for the future development of personalized creative writing tools while highlighting the potential of combining generative AI with node graphs. This opens new research directions for supporting cross-cultural moral reflection.