Confidence Estimation Using Machine Learning in Immersive Learning Environments

As the development of Virtual Reality and Augmented Reality (VR/AR) technology rapidly advances, learning in an artificial immersive environment becomes increasingly feasible. Such emerging technology not only facilitates and promotes an efficient learning process, but also reduces the cost of access to learning materials and environments. Current research mainly focuses on the development of immersive learning environments and the adaptive learning methods based on interactions between trainees and the environment. However, valuable human biometric data available in immersive environments, such as eye gaze and controller pose, have not been explored and utilized to help understand the affective state of the trainees. In this paper, we propose a machine-learning based research framework to estimate trainees' confidence about their decisions in immersive learning environments. Using this framework, we designed an experiment to collect biometric data from a multiple-choice question and answer session in an immersive learning environment. This includes collecting answers from 10 participants on 35 questions and their self-reported confidence in their answers. A Long Short-Term Memory neural network model was used to analyze the data and estimate the confidence with 85.6% accuracy.