Feasibility, Usability, and Effectiveness of a Machine Learning-Based Physical Activity Chatbot: Quasi-Experimental...

Area of Science:

• Digital Health
• Behavioral Science
• Machine Learning

Background:

• eHealth and mobile health interventions show moderate success in increasing physical activity.
• Chatbots offer potential for continuous physical activity monitoring and user engagement.
• Limited studies have evaluated chatbot interventions for physical activity.

Purpose of the Study:

• To investigate the feasibility, usability, and effectiveness of a machine learning-based physical activity chatbot.
• To assess the impact of a chatbot on daily step counts and self-reported physical activity.

Main Methods:

• A quasi-experimental design without a control group was used.
• Participants used a Fitbit for activity tracking and a chatbot via the Messenger app.
• The chatbot provided daily updates, motivational messages, and adaptive goal adjustments.

Main Results:

• Participants increased daily step counts (627 steps/day) and weekly physical activity (154.2 min/week).
• Participants were more likely to meet physical activity guidelines (OR 6.37).
• Usability was rated highly, but chatbot functionality was affected by Facebook policy changes.

Conclusions:

• A machine learning-based chatbot can significantly increase physical activity and is moderately accepted.
• Platform independence is essential for reliable chatbot intervention delivery.
• Future research should focus on independent platforms for chatbot deployment.