Abstract
The growing integration of artificial intelligence into air-combat systems accelerates the collection and fusionof data, yet leaves humans with ultimate responsibility for purely cognitive, high-stakes decisions such as the “shoot / no-shoot” judgment. Because the operator must integrate heterogeneous information within seconds, this decision imposes asubstantial cognitive load. The present study aims to improve performance on this task through a training program builtfrom operational feedback and implemented on a lightweight, easily deployable platform. Seventy fighter-squadron aircrewmembers from the French Air and Space Force participated: an expert group (N = 39) and an intermediate group (N = 31).Intermediates completed a pre-test, a 45-minute training session and a post-test, whereas experts completed only the pre-test. Results on performance (accuracy and response time) show that the training material successfully engages expertknowledge; moreover, intermediates exhibited significant gains after a training session. Transfer of these gains in anoperational setting remains to be demonstrated in future work. Nevertheless, the findings support the development ofsquadron-ready training modules that complement existing instructional tools. Even as AI usage intensifies, humans willremain in the decision loop. Thus, sustained efforts in education and training are essential to keep operators effective, whilealso understanding how to use AI in an appropriate way to support human decision-making, and to ensure that technologicaladvances do not leave people “behind the aircraft.”