Detecting Mind Wandering During Film Viewing
Angela Stewart (University of Notre Dame)
Co-authors: Sidney D’Mello (University of Notre Dame), Nigel Bosch (University of Notre Dame)
A key goal of entertainment applications such as films is to capture and sustain the viewer’s attention. However, involuntary shifts in attention from task-related (the film) to task-unrelated thoughts (called mind wandering) occur quite frequently during film viewing and signal a lack of engagement. We propose that entertainment applications could be improved by detecting and responding to mind wandering. As a first step in this direction, we developed a video-based detector of mind wandering during film viewing. We collected videos of the face and upper bodies of 109 participants who self-reported when they caught themselves mind wandering while watching a 32.5 minute commercial film. We used computer vision techniques to extract facial features and body movements from the videos and trained supervised classifiers to identify periods of mind wandering. Here we describe preliminary results for our models that detect mind wandering at a rate that reflected a 31% improvement over a conservative chance baseline.