Beyond Numbers: Creating Analogies to Enhance Data Comprehension and Communication with Generative AI

Title of the Paper

Beyond Numbers: Creating Analogies to Enhance Data Comprehension and Communication with Generative AI

Paper Information

• Topic Area: Data Visualization, Generative Artificial Intelligence
• Keywords: Data Analogy, Generative AI, Design Space, Visualization Tools, Data Communication, Data Comprehension, Creativity Support Tools, User Studies

Research Background and Problem

• Identified Problems or Challenges:

  1. The abstract and complex nature of numerical data often hinders readers' understanding and perception of the content.
  2. Designing data analogies is time-consuming and requires high-level design skills, making it particularly challenging for non-experts.
  3. Existing automated tools are limited to specific datasets and lack innovation and diversity.
  4. There is still a lack of systematic frameworks for using generative AI to assist in analogy design.

• Why This Problem is Important: Data analogy techniques, by associating abstract data with familiar metrics, can effectively enhance data communication. They have broad application potential in fields such as journalism, educational materials, public science, and marketing.

• Research Motivation and Related Work: The authors explore how generative AI can simplify and improve the efficiency of analogy design while supporting the creation of innovative analogies through design spaces and tools. Related fields include research on analogy functions in cognitive science, data visualization tools, and the application of large language models.

Solution

• Proposed Method or Solution: The authors propose a generative AI-based tool called "AnalogyMate," designed to automatically generate data analogies and related design solutions. The system primarily consists of two modules:

  1. Analogy Design Module: Provides automated data analogy options and revises analogy descriptions.
  2. Graphic Design Module: Generates corresponding visual design solutions and materials.

• Innovations:

  1. Introduced a systematic design space that includes classification criteria such as analogy strategies, metric transformations, and visual presentation layouts.
  2. Utilized GPT-3.5 for analogy design and integrated Stable Diffusion for generating visual materials.
  3. Incorporated user interaction, allowing users to customize key steps and optimize the generation process.

• Implementation Steps:

  1. Generate analogy objects using chain-of-thought methods and guided examples.
  2. Revise analogy objects and numerical descriptions to ensure accuracy and clarity.
  3. Propose visual design solutions, including recommendations for keywords related to mood, color schemes, and style dimensions.
  4. Generate specific graphic materials based on user-selected keywords.

Research Outcomes

• Specific Outcomes:

  1. Developed and implemented a creativity support system, AnalogyMate.
  2. Constructed a design space and collected a dataset of 138 data analogy cases, providing structured guidance for analogy design.
  3. Validated the system's effectiveness through two user studies, showing that the system significantly improved analogy design efficiency and enhanced data comprehension.

• Advantages Compared to Existing Solutions:

  1. Leveraged generative AI to enhance creativity diversity and efficiency.
  2. Integrated full-process support for analogy design and visual material generation, streamlining the design process.
  3. Provided user interaction features, enabling users to adjust design schemes at key steps.

• Experimental or Evaluation Results:

  1. User studies demonstrated that AnalogyMate could generate more analogy design solutions in significantly less time compared to traditional search methods.
  2. Data analogies facilitated participants' understanding, improved their perception of data scale, and increased their interest and engagement with the data context.
  3. CSI evaluations showed that the system performed well in creativity support, achieving a B-grade score.

• Limitations and Future Directions:

  1. The analogy design framework does not fully account for differences in familiarity across cultures and backgrounds. Future work could incorporate knowledge graphs for optimization.
  2. Generative AI occasionally produces inaccurate data or calculation errors, requiring further refinement of the validation process.
  3. The system's personalized recommendations for users remain insufficient, and future work could enhance various personalized design options.
  4. Explore the possibility of directly outputting complete design diagrams to optimize user experience.

Conclusion

This paper introduces the AnalogyMate tool and design space, optimizing the data analogy design process and improving user design efficiency and creative thinking. The study highlights the potential of generative AI in data communication and provides clear directions for future research and applications.