Data-driven group comparisons of eye fixations to dynamic stimuli

Area of Science:

• Cognitive Science
• Neuroscience
• Political Psychology

Background:

• Eye tracking technology has advanced to analyze dynamic video stimuli, moving beyond static images.
• Analyzing eye movement data from videos presents challenges, including labor-intensive region of interest (ROI) coding and the lack of mainstream automated computer vision techniques.
• Existing methods struggle with defining relevant ROIs for video frames, necessitating new approaches for efficient data analysis.

Purpose of the Study:

• To evaluate a novel, easy-to-implement, data-driven method for analyzing eye tracking data from video stimuli.
• To investigate differences in eye movements between politically left-wing and right-wing individuals viewing video clips of politicians.
• To assess the method's capability in predicting group membership and identifying salient stimuli.

Main Methods:

• A data-driven methodology was employed to analyze eye tracking data.
• Participants self-reported their political leaning (left-wing or right-wing).
• Eye movements were recorded while participants watched video clips of left- and right-wing politicians.

Main Results:

• The proposed method accurately predicted group membership based on eye movement patterns.
• The method successfully identified specific video clips that effectively distinguished between political groups.
• Analysis revealed the precise sections within video clips that exhibited the most significant differences in gaze behavior between groups.

Conclusions:

• The developed methodology facilitates the analysis of group differences in gaze behavior during video viewing.
• This approach aids in identifying critical stimuli for future research, such as follow-up studies or applications in saccade diagnosis.
• The findings underscore the potential of data-driven methods to overcome challenges in analyzing complex eye tracking data from dynamic stimuli.