Neural Computational Model Predicts Attentional Dynamics during Immersive Search in Virtual Reality

Abstract

Much of the scientific understanding of visual attention has come from desktop paradigms where body, head, and eye movements are restricted. This stands in contrast to the ability to search and navigate with few constraints in the real world. To bridge this gap, a computational model parametrized on a wide range of well-controlled desktop search tasks was evaluated on its ability to predict search behaviors in a less constrained, immersive environment in virtual reality. A set of validation metrics showed that the models attention map could predict empirical behaviors such as eye gaze and manual responses in VR despite the field-of-view varying from one moment to another from the participants movements. The present work quantifies the real-world applicability of laboratory-based theory and highlights ways to address outstanding limitations in explaining naturalistic visual search behaviors.