Consider the example of a computer science lab doing testing with a relatively sophisticated eye tracking unit (this is a real example of a recent experiment in a lab I visited). The researchers’ main goal is to investigate how students attend to areas of a computer screen in the context of working on a group project. In particular, they want to know what people are looking at when they are presented with large amounts of data and have to select information that is relevant for the project.
 
The researchers plan to use small groups of three students and have one student fitted out with an eye tracker to test multiple user interfaces. By this method, they should be able to find out where the student is looking and when the student is looking there. Accumulating data from a large number of students should provide a good picture of how students interact with large data sets on a screen. The general idea is to find out what is going on and then presumably to change the learning circumstances (or screen organization) so that in future students attend more often to important parts of the data.
 
Up to this point, all appears to be well and the researchers are onto a great idea. However, it turns out that this eye tracking method fails to take into account behavioural aspects of the students that may have a profound impact on where people look. For example, gender differences in behaviour (and possibly brains) of the students are likely to have an important effect. Our own recent research using 1st-person perspective virtual reality shows that males and females navigate differently in virtual screen space and that they may use different strategies to reach the required goal. In fact, we can measure these differences by tracking eye movements and comparing them to behavioural performance.
 
So what does this mean for the computer science study? Well, it would appear that if gender differences aren’t considered, then the results of the computer science study will be muddied.  The researchers might not even find the differences they are looking for.  They will also miss important information that could have an impact on how group projects are organized.  For example, it might turn out that the strategies used by males and females are complementary and could be manipulated for optimal learning. If this were the case, then it would be important to have representatives of both sexes in each group. Or perhaps male/female students could be trained in the use of alternative strategies to optimize their interactions with large data sets on the screen. Or alternatively, it could be necessary to study how gender differences in strategy selection are influenced by how the data is presented. In a field like computer science (where females are now beginning to be well represented) this information would be invaluable but is completely overlooked when only the eye movements of the students are studied.
 
In short, simply recording eye movements using the latest and greatest equipment may not achieve the best result. In the case of the computer science study, useful information is acquired but some extremely important information is never considered. And the computer science study is just one small example of a much larger issue. Given the multiple aspects of human behaviour and given how important ‘screen time’ is in today’s world, it is important for researchers and businesses alike to know what aspects of behaviour influence where people look when they interact with screen space. This is a point that has frequently been overlooked in the application of eye tracking technologies. For optimal results an interdisciplinary approach would seem to be advisable, combining knowledge from the cognitive and behavioural sciences with engineering and technological approaches to eye tracking. Ultimately, knowing more about human behaviour in the context of ‘screen time’ will support the construction of better interfaces that can be more easily optimized.
 
(Author: Sharon Lee)