An Evaluation of the Effect of App-Based Exercise Prescription Using Reinforcement Learning on Satisfaction and...

Area of Science:

• Digital Health
• Exercise Science
• Machine Learning

Background:

• Sedentary lifestyles necessitate innovative public health interventions like smartphone apps for personalized exercise.
• Mobile technologies offer a scalable, cost-effective method for remote exercise program delivery.
• Machine learning (ML), particularly reinforcement learning (RL), can enhance user engagement and program effectiveness through personalization.

Purpose of the Study:

• To evaluate the i80 BPM app's impact on user satisfaction and exercise intensity.
• To assess the effectiveness of ML-generated exercise programs for remote prescription.

Main Methods:

• Participants completed 12 weeks of exercise, alternating between RL-individualized and non-individualized sessions.
• Exercise intensity and user satisfaction were measured using validated scales.
• Randomized crossover design with 62 participants.

Main Results:

• The RL-individualized condition led to significantly higher exercise intensity (mean 5.82 vs. 5.19).
• Participants reported greater satisfaction in the RL condition (mean 4.0 vs. 3.73).
• 62 participants completed 559 sessions over 12 weeks.

Conclusions:

• RL effectively increases exercise intensity and enjoyment, demonstrating ML's potential in remote interventions.
• Personalized exercise prescriptions improve adherence and satisfaction, crucial for long-term fitness.
• App-based, personalized exercise may outperform traditional methods, impacting public health by reducing inactivity.