Dancing with the Avatars: Feedforward Learning from Self-Avatars

Abstract

Feedforward is a training technique where people observe themselves perform a new skill to promote rapid learning, commonly implemented via video self-modelling. Avatars provide a unique opportunity to self-model skills an individual's physical self cannot yet perform. We investigated the use of avatars in video-based learning and explore the potential of feedforward learning from self-avatars. Using modern dancing as a skill to learn, we compared the user experience when learning from a human training video and an avatar training video, considering both self-avatars (n=8) and gender-matched generic avatars (n=8). Our results indicate that learning from avatars can improve the user experience over learning from a human in a video, providing attentional and motivational benefits. Furthermore, self-avatars make the training more relatable and immersive than generic avatars. We discuss the implications from this preliminary work, highlighting methodological considerations for feedforward learning from avatars and promising future work.