Analyzing Eye-Tracking Information in Visualization and Data Space: From Where on the Screen to What on the Screen

Area of Science:

• Human-Computer Interaction
• Data Visualization
• Cognitive Science

Background:

• Current eye-tracking analysis relies on image space, using methods like heatmaps or manually defined areas of interest (AOI).
• These traditional methods require substantial manual effort and are impractical for large-scale or interactive studies.

Purpose of the Study:

• To demonstrate the feasibility and effectiveness of analyzing eye-tracking data in data space.
• To introduce a novel algorithm for translating gaze coordinates into viewed data objects with high accuracy.

Main Methods:

• Instrumenting visualizations with open-source code to map gaze coordinates to data objects in real-time.
• Collecting eye-tracking data from nine users interacting with instrumented visualizations.
• Developing and applying an algorithm to translate gaze data into viewed objects.

Main Results:

• Eye-tracking data collected in data space aligned well with user tasks.
• The data was comparable to annotations produced by human coders.
• The novel algorithm achieved higher accuracy in identifying viewed objects than dynamic AOI binning.

Conclusions:

• Analyzing eye-tracking in data space is a viable and effective alternative to image space analysis.
• This approach enhances the scalability and accuracy of eye-tracking studies, especially for interactive visualizations.
• Further exploration of challenges, opportunities, and benefits of data space eye-tracking analysis is warranted.