Eye tracking cognitive load using pupil diameter and microsaccades with fixed gaze.

Krzysztof Krejtz,Cezary Biele,Andrew T Duchowski,Anna Niedzielska,Izabela Krejtz,Susana Martinez-Conde

doi:10.1371/journal.pone.0203629

Abstract

Pupil diameter and microsaccades are captured by an eye tracker and compared for their suitability as indicators of cognitive load (as beset by task difficulty). Specifically, two metrics are tested in response to task difficulty: (1) the change in pupil diameter with respect to inter- or intra-trial baseline, and (2) the rate and magnitude of microsaccades. Participants performed easy and difficult mental arithmetic tasks while fixating a central target. Inter-trial change in pupil diameter and microsaccade magnitude appear to adequately discriminate task difficulty, and hence cognitive load, if the implied causality can be assumed. This paper’s contribution corroborates previous work concerning microsaccade magnitude and extends this work by directly comparing microsaccade metrics to pupillometric measures. To our knowledge this is the first study to compare the reliability and sensitivity of task-evoked pupillary and microsaccadic measures of cognitive load.

Highlights

Cognitive Load Theory (CLT) [1] plays an important role in Human-Computer Interaction (HCI) research
We give a direct comparison of the measures by testing a multinomial logistic regression (MLR) model for discriminating between task difficulty
We briefly reviewed cognitive load and its connection to task-evoked eye movement measures: pupillary and microsaccadic responses

Summary

Introduction

Cognitive Load Theory (CLT) [1] plays an important role in Human-Computer Interaction (HCI) research. There is a pressing need for a non-invasive measure of individuals’ cognitive load, as it can guide designers of interactive systems to avoid overloading users. Measurement of cognitive load could allow a system to respond appropriately, potentially either by toning down or ramping up the level of task difficulty e.g., as in e-learning systems [2], or by adapting mission-critical systems to the user’s cognitive state [3]. A reliable, and possibly realtime, measurement of cognitive load is highly desirable. Due to a lack of its reliable measurement, only a weak link exists between Human-Computer Interaction and Cognitive Load Theory [8]

Methods

Results

Discussion

Conclusion