Assessing Physiological Signal Utility and Sensor Burden in Estimating Trust,Situation Awareness,and Mental Workload

Abstract

Effective human-autonomy teaming is increasingly important to ensuring mission success in operational environments. Modeling operators’ cognitive states, including trust, situation awareness (SA), and mental workload (WL), may improve human-autonomy team performance by informing autonomous systems about human operators. Subjective questionnaires are often used to measure these states but are obtrusive and impractical for real-world operations. Integrating observable and physiological measures could enable unobtrusive measurements of cognitive states. We created models to estimate trust, SA, and WL using observable, physiological, and operator background information (OBI) measures. We collected data from 15 subjects during a spacecraft docking simulation. We developed a LASSO-based algorithm to select features, generated multivariate regression models, and assessed predictive capabilities. Observable and OBI features combined led to the best performing model, indicating that physiological signals do not add significant predictive power. Simultaneous feature selection of SA and WL yielded performance comparable to that of models fit to a single cognitive state, but did not reduce the number of required physiological sensors. The developed algorithm, use of multiple feature modalities, and simultaneously fitting capability can be leveraged for better estimating human cognitive states for human-autonomy teaming.