Using gaze patterns to study and predict reading struggles due to distraction.

We analyze gaze patterns to study how users in online reading environments cope with visual distraction, and we report gaze markers that identify reading difficulties due to distraction. The amount of visual distraction is varied from none, medium to high by presenting irrelevant graphics beside the reading content in one of 3 conditions: no graphic, static or animated graphics. We find that under highly-distracting conditions, a struggling reader puts more effort into the text -- she takes a longer time to comprehend the text, performs more fixations on the text and frequently revisits previously read content. Furthermore, she reports an unpleasant reading experience. Interestingly, we find that whether the user is distracted and struggles or not can be predicted from gaze patterns alone with up to 80% accuracy and up to 15% better than with non-gaze based features. This suggests that gaze patterns can be used to detect key events such as user strugglefrustration while reading.