GymBuddy: Embedding Motivational Theories in Conversational AI Fitness Coaches

Abstract

Home fitness platforms offering digitally mediated training experiences with virtual instructors or remote peers have gained widespread adoption, relying on screens and immersive setups to support engaging workouts. However, while such platforms provide structured guidance and performance feedback, it remains an open question how to design motivational strategies for AI fitness coaches. To investigate this, we implemented an AI chatbot, GymBuddy, with four different personalities that reflect coaching styles from motivational theories, which are autonomy-supportive, competence-supportive, relatedness-supportive, and controlling coaching. We evaluated the system in a user study (N = 14) where the participants were instructed by the coaches while exercising. Our results indicate that the characteristics of the digital coaches significantly impacted the participants' performance and motivation. The study highlights the potential for AI to democratize fitness coaching and create personalized training experiences to enhance training motivation.