Abstract:

Previous studies have used mouse tracking as a tool to mea- sure usability of webpages, user attention and search rele- vance. In this paper, we go beyond measurement of user behavior to prediction of the resulting user experience from mouse patterns alone. Specifically, we identify mouse mark- ers that can predict user frustration and reading struggles at reasonably high accuracy. We believe that mouse-based pre- diction of user experience is an important advance, and could potentially offer a scalable way to infer user experience on the web. In addition, we demonstrate that mouse tracking could be used for applications such as evaluating content layout and content noticeability; we apply this in particular to advertise- ments. More generally, it could be used to infer user attention in complex webpages containing images, text and varied con- tent, including how attention patterns vary with page layout and user distraction.