Modeling Cognitive Load in Mobile Human Computer Interaction Using Eye Tracking Metrics

Antony Joseph*, J. Sharmila Vaiz, Ramaswami Murugesh

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingsScientificpeer-review

Abstract

Modeling cognitive load of user interaction based on ocular parameters have become a dominant method for exploring usability evaluation of interfaces for systems and applications. Growing importance of Artificial Intelligence in Human Computer Interaction (HCI) has proposed many approaches to understand users’ need and enhance human centric method for interface design. In particular, machine learning-based cognitive modeling, using eye tracking parameters have received more attention in the context of smart devices and applications. In this context, this paper aims to model the estimated cognitive load values for each user into different levels of cognition like very high, high, moderate, low, very low etc., while performing different tasks on a smart phone. The study focuses on the use behavioural measures, ocular parameters along with eight traditional machine learning classification algorithms like Decision Tree, Linear Discriminant Analysis, Random Forest, Support Vector Machine, Naïve Bayes, Neural Network, Fuzzy Rules with Weight Factor and K-Nearest Neighbor to model different levels of estimated cognitive load for each participant. The data set for modeling consisted of 250 records, 11 ocular parameters as prediction variables including age and type of task; and three types of classes (2-class, 3-class, 5-class) for classifying the estimated cognitive load for each participant. We noted that, Age, Fixation Count, Saccade Count, Saccade Rate, Average Pupil Dilation are the most important parameters contributing to modeling the estimated cognitive load levels. Further, we observed that, the Decision Tree algorithm achieved highest accuracy for classifying estimated cognitive load values into 2-class (86.8%), 3-class (74%) and 5-class (62.8%) respectively. Finally, from our study, it may be noted that, machine learning is an effective method for predicting 2-class-based (Low and High) cognitive load levels using ocular parameters. The outcome of the study also provides the fact that ageing affects users’ cognitive workload while performing tasks on smartphone.
Original languageEnglish
Title of host publicationAdvances in Artificial Intelligence, Software and Systems Engineering
Subtitle of host publicationProceedings of the AHFE 2021 Virtual Conferences on Human Factors in Software and Systems Engineering, Artificial Intelligence and Social Computing, and Energy, July 25-29, 2021, USA
EditorsTareq Z. Ahram, Waldemar Karwowski, Jay Kalra
Place of PublicationCham, Switzerland
PublisherSpringer
Pages99-106
Number of pages8
Edition1
ISBN (Electronic)978-3-030-80624-8
ISBN (Print)978-3-030-80623-1
DOIs
Publication statusPublished - 8 Jul 2021
MoE publication typeA4 Conference publication

Publication series

NameLecture Notes in Networks and Systems
PublisherSpringer Cham
Volume271
ISSN (Print)2367-3370
ISSN (Electronic)2367-3389

Fingerprint

Dive into the research topics of 'Modeling Cognitive Load in Mobile Human Computer Interaction Using Eye Tracking Metrics'. Together they form a unique fingerprint.

Cite this