abstract
- Situation awareness (SA)-an operator's perception, comprehension, and projection of goal-critical information-is fundamental to the safety and performance of human operators. Recent advances in autonomous systems can reduce operator SA, so researchers have sought real-time, nondisruptive indicators of SA to enable SA-based adaptive cooperation in human-autonomy teams. However, gold-standard freeze-probe measures of SA are not validated for use as ground truth in real-time predictive models. Working memory constraints force single-trial measures to be partial by nature. Existing workarounds smooth over temporal dynamics, precluding real-time predictive models. This work shows that a 3-trial moving average SA score reduces measurement noise while preserving temporal information. Moving average scores are more strongly correlated (r = 0.36, p < 0.01) with performance than single trial SA scores and can be predicted by single-trial physiological signals with greater accuracy (standardized mean absolute error = 0.61, Q2 = 0.36) than single trial SA scores.