What can machine learning teach us about habit formation? Evidence from exercise and hygiene

Area of Science:

• Behavioral Science
• Machine Learning
• Health Informatics

Background:

• Habit formation is crucial for behavior change but its temporal dynamics are poorly understood.
• Previous research often relies on self-reported data and lacks objective measures of habit formation.
• The concept of a fixed number of days to form a habit is widely popularized but lacks empirical support.

Purpose of the Study:

• To quantitatively characterize habit formation using objective behavioral data and machine learning.
• To investigate the time course of habit formation in distinct behavioral contexts: gym attendance and hospital handwashing.
• To explore the relationship between habit strength, predictability, and intervention responsiveness.

Main Methods:

• Applied a novel machine learning technique, Predicting Context Sensitivity (PCS), to analyze large-scale objective panel data.
• Utilized over 12 million gym attendance records and 40 million hospital handwashing observations.
• Developed individual habit formation curves based on time-series predictability of behavior.

Main Results:

• Habit formation for gym attendance typically requires months, significantly longer than the commonly assumed 'magic number' of days.
• Habit formation for hospital handwashing occurs much faster, often within weeks.
• Individuals with higher behavioral predictability (stronger habits) exhibited reduced responsiveness to interventions promoting attendance.

Conclusions:

• Habit formation is a lengthy process, varying significantly across different behaviors and contexts.
• The study challenges simplistic 'magic number' theories of habit formation, emphasizing individual variability and context.
• Strong habits may confer resistance to external influences, impacting the efficacy of behavioral interventions.