A Review and Collation of Graphical Perception Knowledge for Visualization Recommendation

Paper Title

A Review and Collation of Graphical Perception Knowledge for Visualization Recommendation

Paper Information

• Subject Area: Data Visualization Recommendation and Graphical Perception
• Keywords: Graphical Perception, Data Visualization Design, Visualization Recommendation, User Perception, Visual Encoding, Literature Review

Research Background and Issues

• Identified Problems or Challenges:

  • A low adoption rate of graphical perception research findings in visualization recommendation algorithms.
  • Inconsistencies or contradictions among results from different graphical perception studies.
  • Lack of shared datasets and standardized rules to integrate graphical perception research into visualization recommendation systems.
  • Many visualization recommendation systems rely on only a few graphical perception studies, leading to potentially outdated or limited encoding decisions.

• Significance:

  • Graphical perception results directly impact users' interpretation of data and the effectiveness of visualization design, enabling recommendation algorithms to provide more accurate and user-friendly visualization designs.
  • Helps avoid erroneous or inefficient design decisions, improving users' analytical efficiency and comprehension.

• Research Motivation and Related Work:

  • Transforming graphical perception literature into actionable design guidelines and integrating them into visualization recommendation systems.
  • Previous studies primarily focused on summarizing graphical perception findings or analyzing the behavior of visualization recommendation systems, without systematically exploring the relationship and impact between the two.

Solution

• Proposed Method or Solution:

  • Conduct a review of 59 graphical perception-related papers, including theoretical and experimental studies.
  • Develop a JSON dataset to integrate existing graphical perception knowledge, documenting design comparison experiments and graphical perception rules in detail.
  • Provide specific design guidelines and demonstrate improvements through three representative visualization recommendation systems (Foresight, Voyager, Draco).
  • Share code for automatically converting graphical perception findings into Draco constraints.

• Innovations:

  • Extends beyond ranking individual encodings to explore the design space of multi-encoding combinations, considering the impact of task types and data characteristics.
  • Offers design guidelines tailored to specific tasks and data characteristics, rather than relying solely on broad theoretical principles.
  • The dataset can be directly integrated into recommendation algorithms, enhancing their ability to cater to data-specific and task-specific needs.

• Implementation Steps and Techniques:

  1. Define the visualization design space (e.g., excluding 3D visualizations, network diagrams, and animated interactions).
  2. Select and analyze 59 papers, documenting the impact of visual encodings and graphical designs on user perception and task performance.
  3. Develop a structured schema for data recording, including design types, tasks, and result documentation.
  4. Summarize specific rankings of visual encodings and design guidelines.
  5. Translate the guidelines into hard or soft constraints for visualization recommendation systems.

Research Outcomes

• Specific Outcomes:

  • Created a JSON dataset encompassing graphical perception rules and experimental results, which can be directly utilized by recommendation systems.
  • Summarized guidelines for matching visual encodings with specific task types and data characteristics.
  • Analyzed and synthesized interference effects among encodings from experimental results, tabulating optimal visualization designs for task-related scenarios.

• Comparative Advantages Over Existing Solutions:

  • Integrates a broader range of theoretical and experimental findings into recommendation systems, rather than relying on a limited number of studies.
  • Provides specific design guidelines for recommendation algorithms, enhancing their precision and effectiveness.
  • Demonstrates improvements in existing visualization recommendation systems such as Draco, Voyager, and Foresight.

• Experimental or Evaluation Results:

  • Conducted case studies using three existing recommendation systems, demonstrating that the improved recommendations better align with users' task requirements.
  • Examples show that using improved color or shape encodings leads to more accurate user interpretation of recommended visualizations.

• Limitations and Future Directions:

  • The current dataset lacks experimental coverage of certain encodings (e.g., texture, shape) and some graphical designs.
  • Insufficient reproducibility studies and integration to resolve conflicts among different studies.
  • Limited exploration of graphical aesthetics, user intuition, and personalized preferences.
  • Future work should expand the design space to include 3D visualizations, network graphs, and interactive dynamic designs, while focusing on theoretical advancements and experimental extensions.