The Good Gaze Analysis is based on a sophisticated computer model of human attention.

In its Eyetracking experiments, the Neurobiopsychology group watches hundereds of subjects a year as they view everyday real-world scenes, artificial images, websites and other visual material.
The very same images are shown to our computer model. It will then analyse the images' features, such as color, contrast, shapes, borders, and so on.

In addition, we put in all the information gained by eyetracking: where people looked, for how long, in which order, etc.
From those two things, the model learns which types, combinations and locations of features are actually attracting attention.

When you upload an image to the Good Gaze server, the attention model will analyse it just as before by checking different visual features.
It then uses the knowledge it gained by "watching" humans, to predict how people will look at this image.

Science developed a deep understanding of humans' attentional behaviour, but there is still
a lot left to discover. Specifically, prediction gets harder the more our conscious systems are involved. Seeing an image for the first time, we are primarily guided by the kind of attention Good Gaze is so good in predicting. This helps us in getting the big picture of what we are seeing. On this basis we will then guide our eyemovements depending on our task or aim, e.g. reading through a block of text serially.