Machine Learning Methods to Personalize Persuasive Strategies in mHealth Interventions That Promote Physical...

Area of Science:

• Digital Health
• Behavioral Science
• Machine Learning Applications

Background:

• Physical inactivity is a global health concern, despite the known benefits of physical activity (PA).
• Mobile health interventions show promise for promoting PA, with personalization enhancing their effectiveness.
• Machine learning (ML) offers advanced capabilities for personalizing persuasive strategies in PA interventions, yet a comprehensive overview is lacking.

Purpose of the Study:

• To provide an overview of machine learning (ML) techniques used to personalize persuasive strategies in mobile health interventions for promoting physical activity (PA).
• To present a categorization of these ML techniques to guide future intervention design and development.

Main Methods:

• A scoping review was conducted following Arksey and O'Malley's framework and PRISMA-ScR guidelines.
• Searches were performed in Scopus, Web of Science, and PubMed, with screening using ASReview software.
• Data extraction focused on study details, target groups, PA interventions, technology used, and ML personalization strategies.

Main Results:

• Forty papers from 27 projects were included, categorized by personalization dimension.
• Supervised learning (SL) and reinforcement learning were the primary ML types identified for personalizing message timing and content.
• A three-level categorization emerged: personalization dimension, persuasive strategy, and ML type, detailing specific applications like message timing and content personalization.

Conclusions:

• The developed categorization of persuasive strategies and corresponding ML methods offers valuable insights for designing personalized PA interventions.
• This overview serves as a foundation for future research, potentially expanding the categorization with more detailed ML methods and personalization dimensions.