Why do we need high-fidelity synthetic eye movement data and how should they look like?

Abstract Eye tracking has been a popular behavioral recording method across psychology, neuroscience, and computer science, but a need for large and diverse datasets has emerged. Synthetic eye movement data offer a promising complement, yet it remains unclear which aspects of real oculomotor behavior they must capture. This paper has three objectives: to clarify why synthetic eye movement data are needed, to outline what high-fidelity synthetic signals should look like, and to demonstrate how existing longitudinal datasets and subjective reports can guide their design and validation. We analyzed the motivation for synthetic eye movements and presented a framework of eye movement variance: ocassion-specific or state-specific variance, between-individual variance, pipeline induced variance and noise. Finally, we analyze subjective reports collected alongside the GazeBase dataset, demonstrating some ocassion-specific variance in data and setting requirements for state-free synthetic eye movement signals. Competing Interest Statement The authors have declared no competing interest.