PatternSight: A Perceptual Grouping Effectiveness Assessment Approach for Elements in Charts

Area of Science:

• Data Visualization
• Human-Computer Interaction
• Cognitive Science

Background:

• Chart authoring tools are more accessible, but effective data pattern representation remains challenging.
• Lack of understanding of perceptual mechanisms hinders effective chart design.
• Authors need tools to assess visualization effectiveness and select appropriate representations.

Purpose of the Study:

• To propose a perceptual grouping simulation model for predicting how viewers organize visual elements into graphical patterns.
• To develop a tool that assists chart authors in evaluating and improving visualization designs for pattern representation.

Main Methods:

• Integrating perceptual grouping principles with visual feature extraction of chart elements.
• Developing a simulation model to predict perceptual grouping outcomes.
• Empirical evaluation with human participants to validate model accuracy.
• Embedding the model into a prototype interface called PatternSight.

Main Results:

• The perceptual grouping simulation model accurately learns viewer tendencies and accommodates diverse perceptual outcomes.
• Empirical evaluations confirmed the model's effectiveness in predicting perceptual grouping.
• User experiments showed PatternSight successfully assists authors in optimizing chart designs.

Conclusions:

• The proposed model and PatternSight tool effectively address the challenge of designing visualizations for clear data pattern representation.
• PatternSight empowers chart authors to assess design effectiveness and make informed improvements.
• This work enhances the creation of data visualizations that align with human perceptual mechanisms.