How are Machine Learning and Artificial Intelligence Used in Digital Behavior Change Interventions? A Scoping Review

Area of Science:

• Digital Health
• Behavioral Science
• Artificial Intelligence
• Machine Learning

Background:

• Digital behavior change interventions (DBCIs) increasingly incorporate machine learning (ML) and artificial intelligence (AI).
• Assessing the current real-world applications of ML/AI in DBCIs influencing patient and consumer health behaviors is crucial.
• Understanding the types of ML/AI used and their effectiveness in behavior change is an active area of research.

Purpose of the Study:

• To conduct a scoping review of live DBCIs utilizing ML or AI to influence real-world health behaviors.
• To identify the behavioral domains, target behaviors, and specific ML/AI functionalities employed.
• To evaluate the research quality and limitations associated with these ML/AI-driven DBCIs.

Main Methods:

• A comprehensive scoping review was performed across EMBASE, PsycInfo, PsycNet, PubMed, and Web of Science databases.
• Search terms included ML/AI, behavioral science, and digital health, focusing on live DBCIs.
• Thirty-two articles met the inclusion criteria, with data extracted on behavioral domains, target behaviors, ML/AI types and purposes, and research evaluations.

Main Results:

• Twenty-three DBCIs were found to use AI for influencing real-world health behaviors.
• Cardiometabolic health (21.7%) and lifestyle interventions (17.4%) were the most common domains.
• Classical ML algorithms (43.5%), reinforcement learning (34.8%), and natural language understanding (34.8%) were prevalent ML/AI types.

Conclusions:

• AI in DBCIs shows promise for managing complex data and providing personalized support for behavior change.
• Evidence suggests positive outcomes but highlights limitations including lack of causal detection, low generalizability, and insufficient long-term data.
• Standardizing terminology and enhancing the understanding of ML/AI applications in DBCIs are key opportunities for future research and development.