József Arató

Research

• Currently, I use computer vision models and gamified human experiments to study the visual learning dynamics in complex visual environments. I am interested in the extent to which low level perceptual representations learned by human experts resemble those of computer vision models.
• Eye movements in learning: My PhD focused on how visual statistical learning influences eye movements in a gaze contingent paradigm (Arato, Rothkopf & Fiser, 2024). We found that while implicit learning can proceed even without any effect on eye movements, after sufficient learning or using explicit instruction a tight link emerges between eye movements and performance on a recognition test.
• Time series modelling: I use hidden Markov models to study human gaze dynamics in free viewing and to quantify scanpath similarity. I also use HMMs to study dynamic functional connectivity using large scale fMRI datasets of neurodivergence.
• Birdsong phylogenetics: I study cross-species bird vocalization datasets to study the extent to which evolutionary history can explain their similarity. (Arato & Fitch, 2021)

Students (co)supervised

PhD Students

• Xingyu Long — Pupillary Dynamics and Ambient-Focal Oscillations in Aesthetic Perception

Master Students

• Jannis Bressgott — Dynamic functional connectivity of the reward network in autism: a model comparison
• Nikita Podolin-Danner — Beyond mere presence: The role of social evaluation in working memory performance
• Anja Stojkovic — Understanding Viewer Engagement with Art: A Study of Visual Attention Using Eye Tracking and Machine Learning
• Nicoló Filipucci — Eye-Movement Similarity: Scanpath Analysis for Recognizing Learning Disorders

Code

Selected code repositories:

• Bayesian cognitive science book — Python translation
• Eye movement analysis package
• Birdsong phylogenetic analysis code (Arato & Fitch 2021)