Model-based eye-tracking: a new window to understand individual differences and psychiatric disorders

Our eyes are constantly moving, and where we look reveals what we attend to, influences the decisions we make, and what we remember. While traditional laboratory-based eye-tracking protocols have generated a large body of findings based on well-controlled stimuli, modern methods now offer the ability to collect gaze data at much larger scale and with more naturalistic stimuli. Powerful computational tools also enable new analyses of high-dimensional data, incorporating feature annotation of the stimuli and model-based evaluation of gaze. These advances in both data collection and analysis are providing new insights into individual differences in both health and disease. Here we discuss four key approaches to modeling eye movement data: saliency-based attention phenotyping, data-driven gaze pattern identification, supervised machine learning classification, and unsupervised clustering. We highlight their advantages in psychiatry research, as they inform better understanding of visual attention, provide more fine-grained characterization of individual differences, and make more powerful clinical predictions. Finally, we address key methodological considerations in applying the methods and take stock of future opportunities on the horizon.