# Discriminative Capabilities of Eye Gaze Measures for Cognitive Load Evaluation in a Driving Simulation Task

## TL;DR

This study compares eye tracking and self-report methods to assess cognitive load in driving simulations, finding that eye gaze metrics are effective indicators.

## Contribution

The study introduces eye gaze entropy measures as novel indicators for evaluating cognitive load dynamics during driving.

## Key findings

- Eye gaze entropy measures achieved an average accuracy of 0.95 in distinguishing cognitive load between urban and highway driving.
- Eye gaze metrics showed significant correlations with self-reported cognitive load in urban driving but not in highway driving.
- Combining eye gaze metrics with self-reports can improve cognitive load evaluation and driver state prediction.

## Abstract

Driving is a cognitively demanding task engaging attentional effort and working memory resources, which increases cognitive load. The aim of this study was to evaluate the discriminative capabilities of an objective eye tracking method in comparison to a subjective self-report scale (the NASA–Task Load Index) in distinguishing cognitive load levels during driving. Participants (N = 685) performed highway and urban driving in a fixed-base driving simulator. The N-Back test was used as a secondary task to increase cognitive load. In line with previous studies, the NASA–Task Load Index was shown to be an accurate self-report tool in distinguishing conditions with higher and lower levels of cognitive load due to the additional N-Back task, with best average accuracy of 0.81 within the highway driving scenario. Eye gaze metrics worked best when differentiating between stages of highway and urban driving, with an average accuracy of 0.82. Eye gaze entropy measures were the best indicators for cognitive load dynamics, with average accuracy reaching 0.95 for gaze transition entropy in the urban vs. highway comparison. Eye gaze metrics showed significant correlations with the NASA–Task Load Index results in urban driving stages, but not in highway driving. The results demonstrate that eye gaze metrics can be used in combination with self-reports for developing algorithms of cognitive load evaluation and reliable driver state prediction in different road conditions.