The intervention for the AI-Based Companion group involved interacting with the BodyPark ATOM system (BodyPark Inc., Hong Kong, China), a portable AI fitness companion device. This device is equipped with a $160^{\circ}$ ultra-wide camera and features real-time motion analysis capabilities. During the 20-minute functional circuit training session, the system served as the digital coaching interface, providing automated real-time recognition of the targeted movements, verbal encouragement, and posture corrections.

Emotional Resonance (ER) measures the extent to which participants perceive empathy, encouragement, and emotional companionship during their exercise session. This dimension evaluates the responsiveness of the coach or system to signs of participant exhaustion or form breakdown, as well as the perceived sincerity of the feedback provided, which helps to alleviate exercise-related boredom or anxiety. Theoretically, it is grounded in parasocial interaction and social presence theories, suggesting that users naturally attribute social characteristics to interactive entities, fulfilling fundamental psychological needs for relatedness and competence.

Interactive Dynamics (InD) represents the responsiveness, bidirectional communication, and real-time adaptability of the interaction process between the participant and the coach or system. It evaluates the fluidity of the communication loop, including the speed of movement correction, the naturalness of the exchange, and the system's ability to dynamically adjust guidance based on real-time performance. This construct is derived from human-computer interaction (HCI) theories, emphasizing that a seamless, low-latency feedback loop is essential for maintaining motor-cognitive immersion and optimizing exercise strategies.

Embodied Presence (EP) reflects the degree to which a participant perceives an instructing agent as occupying the same physical or virtual space, facilitating bodily awareness and spatial immersion. This dimension emphasizes spatial perception, physical occupancy, and proprioceptive mapping, allowing participants to intuitively map external postural cues (such as joint angles or body alignment) onto their own internal body schema. It is grounded in cognitive psychology and spatial cognition frameworks, measuring how well the instruction facilitates an intuitive spatial-motor translation.