Assessment of Mental Workload: A Comparison of Machine Learning Methods and Subjective Assessment Techniques

Abstract

Mental workload (MWL) measurement is a complex multidisciplinary research field. In the last 50 years of research endeavour, MWL measurement has mainly produced theory-driven models. Some of the reasons for justifying this trend includes the omnipresent uncertainty about how to define the construct of MWL and the limited use of data-driven research methodologies. This work presents novel research focused on the investigation of the capability of a selection of supervised Machine Learning (ML) classification techniques to produce data-driven computational models of MWL for the prediction of objective performance. These are then compared to two state-of-the-art subjective techniques for the assessment of MWL, namely the NASA Task Load Index and the Workload Profile, through an analysis of their concurrent and convergent validity. Findings show that the data-driven models generally tend to outperform the two baseline selected techniques.