Unpacking Visual Metaphors in Infographics: A Design Space

Paper Title

Unpacking Visual Metaphors in Infographics: A Design Space

Publication Info

• Topic area: Visual metaphor design in infographics
• Keywords: Visual metaphors, infographics, design space, generative models, metaphor ideation, data visualization, cognitive linguistics, reconstruction strategy, source properties, user study

Background and Problem

• Problem / challenge: Designing effective visual metaphors in infographics is challenging, especially for novice designers, due to the complexity of selecting suitable source concepts and devising reconstruction strategies. Existing research lacks systematic guidance for ideating and applying visual metaphors.
• Significance: Visual metaphors enhance data understanding, retention, and emotional engagement, making them critical for effective communication in infographics.
• Motivation and related work: Prior studies have explored metaphor types, automated tools, and guidelines but have not systematically captured the interaction between target data, source properties, and reconstruction strategies. This paper addresses this gap by introducing a structured design space.

Solution

• Proposed approach: A design space for visual metaphors in infographics, characterized by three dimensions: target, source, and reconstruction strategy. The design space is operationalized into actionable knowledge for generative models.
• Novelty:
  1. A dataset of 2,029 metaphoric infographics collected from diverse sources.
  2. A structured design space categorizing visual metaphors across target, source, and reconstruction strategy dimensions.
  3. A design-space-augmented prompting method for generative models to ideate metaphoric designs.
• Procedure and key techniques:
  • Data collection: Gathered 48,552 infographic candidates from curated datasets, online sources, and publications, filtered to 2,029 metaphoric infographics.
  • Coding: Analyzed infographics using thematic analysis to identify target insights, source properties (denotation and connotation), and reconstruction strategies.
  • Evaluation: Conducted a user study with 30 participants comparing design-space-augmented prompts to direct prompts, assessing novelty, diversity, and satisfaction.

Results

• Concrete findings:
  • Design-space-augmented prompting produced infographics ranked higher in novelty and effectiveness (average rank: 5.48 vs. 7.52 for baseline; p < .001).
  • Diversity scores: 5.82 for augmented method vs. 4.96 for baseline (p < .001).
  • Satisfaction scores: 5.34 for augmented method vs. 4.16 for baseline (p < .001).
• Advantage over baselines: The augmented method consistently generated more diverse and satisfactory metaphor designs across all ten data insights, achieving top ranks in eight out of ten insight types.
• Experiments / evaluation:
  • User study with 30 visualization designers, covering ten data insights (e.g., distribution, trend, rank).
  • Tasks included ranking individual infographics and rating diversity and satisfaction of method outputs.
  • Follow-up interviews with six participants provided qualitative feedback.
• Limitations and future work:
  • Current evaluation focuses on designer preferences, not end-user comprehension or retention.
  • Dataset skewed toward Western-centric infographics; future work should expand cultural and linguistic diversity.
  • Ethical considerations for data accuracy and cultural bias in generated infographics.

Summary

This paper introduces a structured design space for visual metaphors in infographics, characterized by target, source, and reconstruction strategy dimensions, derived from a dataset of 2,029 metaphoric infographics. The design space is operationalized into actionable knowledge for generative models, enabling systematic metaphor ideation. A user study demonstrates that design-space-augmented prompting produces more diverse, novel, and satisfactory metaphor designs compared to direct prompting. Future work includes expanding dataset diversity, evaluating end-user comprehension, and developing human-AI co-creative tools for infographic design.