Close

Presentation

Interpretable Models for Near-Real-Time Prediction of Team Cognitive Workload in Complex Sociotechnical Environments
DescriptionThis work develops interpretable models to predict near-real-time cognitive workload (CWL) in teams operating in complex, high-stakes environments. Existing approaches using neurological sensors like EEG are impractical for field use. The proposed approach integrates multimodal data from non-invasive behavioral and physiological sensors to robustly detect CWL changes. It applies multidimensional recurrence quantification analysis (MdRQA) with a novel pattern analysis on time-series data to identify recurring multimodal signatures indicative of different CWL states. An extensive multiparty dataset with EEG, fNIRS, behavioral, and physiological measures from teams performing shared tasks is used. The MdRQA models aim to extract the most predictive multichannel patterns associated with individual and team CWL derived from non-invasive sensing. The resulting interpretable models pinpoint specific multimodal patterns signaling workload transitions. This can enable timely interventions by intelligent systems to optimally manage team CWL and enhance human-machine teaming in demanding environments.
Event Type
Lecture
TimeTuesday, September 10th10:05am - 10:25am MST
LocationFlagstaff
Tracks
Human AI Robot Teaming (AI)