When a voice agent becomes a flight instructor: Effects of instructor and expertise in simulated flight training

Voice agents (VAs) are increasingly used in learning, training, and human-AI interaction. In this study, a VA instructor for simulated flight training was designed to examine training performance and experience compared with a human instructor. The effects of learners' expertise in flight knowledge and simulated flight experience were also investigated. Sixty-eight participants took part in a between-subjects experiment to learn simulated flight in an aircraft simulator with different instructor types. Subjective ratings, objective flight data, and physiological data were collected. The results showed that participants reported positive attitudes toward learning with the VA. Learner expertise influenced mental model development and subjective task load, with participants of higher expertise demonstrating more complete mental models and lower task load. Instructor type was also associated with learning experience and flight behavior. Compared with the human instructor, VA instruction showed directional patterns toward lower subjective task load and higher flight velocity, with comparable overall training outcomes. These findings support the feasibility of VA-based instruction in simulated flight training and highlight its potential as an instructional agent in complex human-computer interaction environments.