Title: Human-AI Interaction Dynamics with Voice-Driven Autonomous Pilots in MEDEVAC Operations
While autonomous systems advance, human involvement remains crucial in critical domains like medical evacuation, where precise human discernment is essential. In our study, we explored human-AI collaboration between an AI Pilot (AP) and a medical professional in simulated MEDEVAC situations. We recruited 22 primarily licensed medical professionals with limited AI or aviation experience who managed patient care and communicated with the AP using voice commands. Participants faced medical and aviation emergencies in four increasingly complex scenarios in our realistic flight simulator. They engaged in bi-directional communication with the AP to make decisions, execute commands, and respond to requests. Metrics included NASA-TLX and SART scores, task performance, trust measures (iTHAu and Merritt), and voice feature usage.
The results demonstrated an effective interaction with the AP. As expected, as scenario complexity increased, workload increased, while situational awareness declined. A general upward trend in trust in AP was observed, with an anticipated decline during scenarios where the AP provided suboptimal suggestions. The voice feature was predominantly employed during situations of uncertainty rather than higher workload situations.
In conclusion, this study demonstrates that human-AI collaboration is feasible but further research is needed to understand the subtle aspects of these interactions in dynamic, safety-critical settings.