Measuring pilot trust in tactical autonomy through the Airborne Trust Assessment Slider

The potential benefits of artificial intelligence and manned-unmanned teaming in aviation have been a staple for research in today’s commercial and military industries. Specific studies on the performance of agents in tactical autonomy - and their effectiveness in augmenting their human counterpart - push the envelope on military capabilities and the future of aerial combat. Leading assertations by those studying this possibility agree that pilot trust is a crucial factor in creating a usable AI client, however, accurately quantifying a unit of trust is an elusive task.

Researchers from the University of Iowa have devised a system for quantifying pilot trust in onboard (local) tactical autonomy during dynamic flight maneuvers flown by a simulated (safety pilot flown) or real AI agent. Through years of testing, real time risk measures from a series of flight state and geometric data were analyzed to not only validate their findings, but discover levels of over-trust, under-trust, and moments where the pilots trust was correctly calibrated.

This thesis is built from data provided by a United States Air Force (USAF) Test Pilot School (TPS) funded staff Test Management Project (TMP) called Have Fortitude and will allow future studies to collect trust assessments via an analog slider input device that an Evaluation Pilot (EP) can leverage to record subjective, real-time, continuous trust ratings.