The growing ubiquity of eye trackers embedded into devices and the ability to infer behavior and cognitive processes using eye tracking data will give rise to a new class of intelligent systems which can adapt to users. However, there remain substantial challenges to using eye tracking data like pupil size to infer cognitive processes due to the relative low signal to noise ratio. For example, the influence of light on pupil size is much larger than the influence of cognitive processes. Thus, these technologies must be able to function in a wide variety of real-world fluctuating lighting conditions. Because of these data challenges models that can predict pupil size and account for non-cognitive influences are needed to ensure reliable estimation of cognitive processes. In this project students will aid in collection of training data, develop various models suitable for predicting pupil size and test how their models generalize to more complex naturalistic data sets.
