Search This Blog

Saturday, November 7, 2009

Eye Tracking Experiment - Can Regions of Visual Attention on a Web Page Be Predicted?

By Michelle Fox and Jia Tan

Researchers from the German DFKI research institute and Microsoft Research department conducted a study using pupil tracking technology to identify the most significant regions of a web page. The study had two objectives: 1) to identify the influence of tasks on the pattern of gaze across a variety of web pages, and 2) to create a predictive model of which elements are most likely to be looked at on a web page. Using a Tobii x50 gaze tracking device paired with an LCD monitor, the researchers logged gaze data of 20 participants on 361 different web pages. Three measures were computed using the visual tracking data: median fixation impact, viewing frequency, and median time of first fixation.

The first goal was to determine where web page viewers focus their visual attention and whether the task they were performing (either page recognition or information searching) effected how they viewed the web page. The outcome showed that for both tasks, page viewers fixate on the top left-hand side of the page. This suggests that web page viewers have been conditioned to believe that the right side of most pages has little relevant content or information, so they quickly scan the left side for clues about content.

Using the data from the original experiment, the researchers approached their second objective. A predictive method was created for finding the most important elements for recognizing a page; however, the model proved too simple to avoid bias for elements on the upper left side of the page. Nevertheless, this predictive model will be used to create compact representations of websites using the most salient regions of the page to aid in the task of page recognition or re-finding.

Using technology that tracks the eyes allows researchers to increase their understanding of users' web-viewing behavior. User interfaces that frustrates the user can easily be identified. It can either be removed or modified to make its function clearer to the user. This information helps web developers, advertisers, etc. to improve page design and create new types of web user interfaces that are pleasing and intuitive. The fundamental take away from the research is the understanding that predicting or expecting users to use an user interface as designed is a flawed approach in the development process. Testing user interactions with the user interface is a necessity in today's competitive business world.

No comments: