Quantifying the Predictability of Visual Scanpaths Using Active Information Storage.

Patricia Wollstadt,Christiane B Wiebel-Herboth,Martina Hasenjäger

doi:10.3390/e23020167

Abstract

Entropy-based measures are an important tool for studying human gaze behavior under various conditions. In particular, gaze transition entropy (GTE) is a popular method to quantify the predictability of a visual scanpath as the entropy of transitions between fixations and has been shown to correlate with changes in task demand or changes in observer state. Measuring scanpath predictability is thus a promising approach to identifying viewers’ cognitive states in behavioral experiments or gaze-based applications. However, GTE does not account for temporal dependencies beyond two consecutive fixations and may thus underestimate the actual predictability of the current fixation given past gaze behavior. Instead, we propose to quantify scanpath predictability by estimating the active information storage (AIS), which can account for dependencies spanning multiple fixations. AIS is calculated as the mutual information between a processes’ multivariate past state and its next value. It is thus able to measure how much information a sequence of past fixations provides about the next fixation, hence covering a longer temporal horizon. Applying the proposed approach, we were able to distinguish between induced observer states based on estimated AIS, providing first evidence that AIS may be used in the inference of user states to improve human–machine interaction.

Highlights

The analysis of visual scanning behavior provides a rich source of information in the investigation of observers’ cognitive processes or states [1,2,3,4,5]
We here propose a novel approach to the information-theoretic analysis of scanpath data, which is able to measure predictability in a scanpath while accounting for temporal dependencies of arbitrary order: we propose to estimate active information storage (AIS) [35] from scanpaths, which measures the predictability of a sequence as the mutual information between the sequences’ past and its state
Our results demonstrate that AIS can be used successfully to decode our experimental condition from scanpath predictability, outperforming the classification based on the gaze transition entropy (GTE), indicating that AIS captures more relevant aspects of the task-induced changes in observer state as encoded in eye movement behavior

Summary

Introduction

The analysis of visual scanning behavior provides a rich source of information in the investigation of observers’ cognitive processes or states [1,2,3,4,5]. One aspect of visual scanning behavior are scanpaths, which denote sequences of consecutive fixations, where a fixation is a period of little to no eye movement that allows for gathering of visual information [6]. Information-theoretic measures have become a popular tool for studying cognitive function through the analysis of human gaze behavior [5,8,12,13,14,15,16,17]. GTE considers sequences of fixations, so-called scanpaths, under the assumption that scanpaths can be modeled as Markov chains of order one, and is calculated as the entropy of the transitions between two consecutive fixations. GTE has been applied in various studies (see [5] for a review), which have shown that changes in GTE are associated with higher task demand [13,19,20], increased anxiety [21,22,23], or sleep deprivation [14]

Objectives

Methods

Results

Discussion

Conclusion