Theories across the social sciences propose mechanistic explanations for health behavior change. Yet, these are often criticized for lack of real-world predictability. In turn, machine learning efforts predominantly focus on prediction. Yet, these are often criticized for lack of interpretability.

In this project, we aim to bridge insights from social science and machine learning to build behavior change models that maximize theory-driven explanatory as well as out-of-sample predictive power.

Collaborators:

  • Danielle Cosme, PhD, University of Pennsylvania
  • Zach Boyd, PhD, Brigham Young University
  • Alice Schwartz, PhD, Dartmouth University