Characterizing Smoking and Drinking Abstinence from Social Media

Area of Science:

• Computational Social Science
• Digital Health
• Addiction Research

Background:

• Social media platforms contain valuable health and well-being signals.
• Understanding long-term abstinence from addiction is crucial but challenged by limited longitudinal data.
• Previous research explored the psychology of addiction but lacked robust methods for tracking long-term recovery.

Purpose of the Study:

• To investigate social media's potential in characterizing and understanding tobacco and alcohol abstinence.
• To develop statistical models for distinguishing long-term from short-term abstinence using online user data.
• To explore the role of social media in supporting positive behavior change and therapeutic interventions.

Main Methods:

• Leveraged nearly eight years of user activity from Reddit communities (StopSmoking and StopDrinking).
• Utilized self-reported 'badge' information from nearly a thousand users as a gold standard for abstinence status.
• Built supervised learning models incorporating linguistic features of user content and social interaction network structures.

Main Results:

• Achieved 85% accuracy in distinguishing long-term abstinence (~one year) from short-term abstinence (~40 days).
• Demonstrated that language patterns and social interactions on social media provide significant cues for addiction outcomes.
• Validated the utility of social media data for tracking and understanding recovery trajectories.

Conclusions:

• Social media data, particularly linguistic and network features, can effectively characterize long-term addiction abstinence.
• Findings highlight the potential of social media as a tool for addiction research and monitoring recovery.
• Social media platforms can serve as valuable spaces for positive behavior change, therapy, and support.