RatioLogo
Back

AI Predicts Your Habits Better Than You Can

New algorithms can now map your hidden habit strength more accurately than traditional methods.

A recent study reveals that computer models can predict toothbrushing behavior with greater precision than self-reports or past actions.

The researchers aimed to answer a significant question: Can we enhance the prediction of human behavior by understanding the intricacies of habit formation? They specifically investigated the role of digital tools, known as Behavior Change Support Systems (BCSS), in fostering healthy habits.

How the Study Was Conducted

  1. Tracking Method: The team tracked toothbrushing habits across two studies.
  2. Participants:
    • Both groups brushed their teeth twice daily for approximately three weeks.
    • Study 1: 40 university students.
    • Study 2: 79 adults.
  3. Sensor Technology: Participants wore tiny Axivity AX3 accelerometers on their toothbrushes, which recorded every brushing session.
  4. Data Collection: Alongside sensor data, participants completed surveys about their habits and attitudes.

The Power of Computer Models

The super-smart computer models, built upon theories of habit growth, significantly outperformed older methods.

In the first study (40 university students), the new model achieved an AUC (Area Under the Curve) of 0.737.

In the second, larger study (79 adults), its AUC was even better at 0.815.

Traditional survey-based models and even past behavior alone could not keep pace with the predictive power of the new algorithmic approach.

Why This Matters

According to the authors:

"The theory-based computational approach would be considered valuable if it led to models that performed better than models based simply on past behavior or on weekly self-reported variables."

This implies that understanding the underlying mechanisms of habit formation helped computers become better predictors of behavior.

Implications for Personalized Support

This breakthrough could lead to BCSS applications that are significantly smarter and more personalized.

Imagine an app that accurately predicts when you might skip brushing and nudges you at precisely the right moment.

The models even identified optimal parameters for habit strengthening and decay:

  • Habit decay parameter: between 0.15 and 0.2
  • Habit gain parameter: between 0.1 and 0.2

These findings are akin to discovering the ideal "ingredients" for a habit recipe.

Future Directions and Challenges

However, these advanced models require fine-tuning for different scenarios and user groups.

  • The study was limited to toothbrushing, necessitating further research for other habits.
  • Future work will focus on making these models adaptable across various behaviors and users without constant adjustments.

Understanding the deep currents of habit formation holds the key to developing powerful new ways to support healthy lives.

Reference

Zhang, C., Vanschoren, J., van Wissen, A., Lakens, D., de Ruyter, B., & IJsselsteijn, W. A. (2021). Theory-based habit modeling for enhancing behavior prediction: A preprint. arXiv:2101.01637v1 [cs.AI].