Operator-agnostic and real-time usable psychophysiological models of trust, workload, and situation awareness

Trust, mental workload, and situation awareness (TWSA) are cognitive states important to human performance and human-autonomy teaming. Individual and team performance may be improved if operators can maintain ideal levels of TWSA. Predictions of operator TWSA can inform adaptive autonomy and resource allocation in teams, helping achieve this goal. Current approaches of estimating TWSA, such as questionnaires or behavioral measures, are obtrusive, task-specific, or cannot be used in real-time. Psychophysiological modeling has the potential to overcome these limitations, but prior work is limited in operational feasibility. To help address this gap, we develop psychophysiological models that can be used in real time and that do not rely on operator-specific background information. We assess the impacts of these constraints on the models' performance. Participants (n = 10) performed a human-autonomy teaming task in which they monitored a simulated spacecraft habitat. Regression models using LASSO-based feature selection were fit with an emphasis on model stability and generalizability. We demonstrate functional model fit (Adjusted R2: T = 0.67, W = 0.60, SA = 0.85). Furthermore, model performance extends to predictive ability, assessed via leave-one-participant-out cross validation (Q2: T = 0.58, W = 0.46, SA = 0.74). This study evaluates model performance to help establish the viability of real-time, operator-agnostic models of TWSA.