The dynamical law behind eye movements: distinguishing between L\'evy and intermittent strategies

TL;DR

This study analyzes eye movement patterns during visual foraging, demonstrating that intermittent search models better fit gaze trajectories than Le9vy walk models, using a combination of analytical, statistical, and machine learning methods.

Contribution

It resolves the Le9vy-intermittent dichotomy in eye movement modeling by providing a method to compare models and showing intermittent search models outperform Le9vy walks in fitting gaze data.

Findings

Intermittent search models better fit eye-gaze trajectories.

Analytical and machine learning techniques effectively compare models.

Eye movements during visual foraging are better described by intermittent strategies.

Abstract

Foraging is a complex spatio-temporal process which is often described with stochastic models. Two particular ones, L\'evy walks (LWs) and intermittent search (IS), became popular in this context. Researchers from the two communities, each advocating for either L\'evy or intermittent approach, independently analyzed foraging patterns and reported agreement between empirical data and the model they used. We resolve this L\'evy-intermittent dichotomy for eye-gaze trajectories collected in a series of experiments designed to stimulate free foraging for visual information. By combining analytical results, statistical quantifiers, and basic machine learning techniques, we devise a method to score the performance of the models when they are used to approximate an individual gaze trajectory. Our analysis indicates that the intermittent search model consistently yields higher scores and thus approximates the majority of the eye-gaze trajectories better.