Fig. 2: Data preprocessing pipeline.

a Participants’ response to “meaningful events” via keyboard button press was used to reduce the data and isolate sections of gaze data based salient time points and action phases. b Button press data corresponded to one of six different types of action phases. This was used to quantify which action phase was more salient to participants (see Fig. 2 for action phase saliency). Specific time points in the video were also isolated based on the frequency of button presses within a given time point. c A sample histogram using Post 2 button press data is shown with time on the x-axis and frequency of button presses on the y-axis. Clear peaks represent time points with the highest participant response. d Action phases and time points from button press data was used to pinpoint specific action phases and time points in gaze data for e gaze analysis. Performance measures, specifically model score, were assessed during time points around their respective assessments.