Detection and Transition Analysis of Engagement and Affect in a Simulation-Based Combat Medic Training Environment

Developing intelligent tutoring systems that respond effectively to trainee or student affect is a key part of education, particularly in domains where learning to regulate one’s emotion is key. Effective affect response relies upon effective affect detection. This paper discusses an upcoming cooperative study between the Army Research Laboratory, Teachers College, Columbia University, and North Carolina State University, with the goal of developing automated detectors that can infer trainee affect as trainees learn by interacting with the vMedic system, which trains learners in combat medicine. In this project, trainee interactions with vMedic will be synchronized with observations of engagement and affect; and physical sensor data on learners, obtained through GIFT’s Sensor Module. The result will be models of trainee affect, ready for integration into the GIFT platform, which can leverage sensor information when available, but which can make reasonably accurate inference even without sensor data.