Evaluating Stress Manipulations in Scenario-Based Command and Control Training: A Multimodal Neural Network Approach

Abstract

Command and Control (C2) in crisis response requires teams to coordinate, process information, make decisions, and take appropriate actions under stress. To effectively prepare C2 teams, scenario-based training is a core instructional approach to train and simulate these real-world pressures by manipulating stress. However, it remains unclear if participants actually experience and respond to these scenarios as intended and thus, whether the required capabilities and skills are effectively trained. In contrast to predominantly qualitative evaluations of scenario design, this study adopts a quantitative, multimodal approach using continuous measures of stress and coordination to evaluate and predict three escalating domestic violence arrest training scenarios: low-, medium-, and high-stress. Physiological stress was indexed via heart rate variability, and team coordination behaviors were systematically video-coded. We use recurrent neural network models to compare physiological-only, behavioral-only, and combined inputs to classify the scenario conditions, thereby advancing more evidence-based scenario design and evaluation of C2 training.