Detecting expertise in decision making under pressure: a virtual reality assessment environment and empirical evaluation

Professions such as military, aviation, submarine operation, and emergency response require individuals to navigate complex environments characterized by limited information, stringent time constraints, and significant pressures. Effective decision making under pressure is crucial in safety-critical professions, yet measuring this expertise remains challenging. Inspired by the military context, this article introduces the virtual reality decision-making expertise (VR-DMX) environment, designed to evaluate decision-making expertise under time constraints within a virtual reality scenario. VR-DMX simulates an amusement arcade where users must decide how to allocate time across various games to maximize ticket earnings. Through two validation studies (N = 60 and N = 76), we examined two metrics: Total Tickets (measuring overall performance) and DMX score (isolating decision-making quality). Both metrics demonstrated symmetrical distributions without floor or ceiling effects, with coefficients of variation comparable to established individual difference measures (32.4-37.4% for Total Tickets; 20.8-27.6% for DMX score). The moderate correlation between metrics (meta-analysis r = 0.771, 95% CI [0.599, 0.943]) indicates they measure related but distinct constructs. Our findings indicate that VR-DMX effectively differentiates individual performance levels and captures a distinct decision-making component that is separate from general cognitive abilities. Comparing decision-making expertise between professionals in safety-critical fields with those without such experience would be a sensible next step to help validate the potential for selection and training applications. VR-DMX was designed to measure decision-making expertise in safety-critical contexts, and initial validation data demonstrating effective differentiation of individual performance levels suggest that continued development could fulfill this design intention for applications in selection, training, and performance prediction.