Abstract

When individuals post to social media or use wearable devices, data generated through these everyday interactions with technology reveal a great deal about behaviors that influence health in ways that were previously not observable. In my work, I seek to leverage this data to characterize and measure the naturalistic manifestations a.k.a digital phenotyping of mental and physical health.

In this talk, we will look at a) uncovering linguistic markers of ADHD using self-declared statuses on Twitter, b) scaling language-based user-level questionnaire-estimated psychological stress predictions to communities, and c) forecasting healthcare utilization as documented in the medical records of a sample of patients using their Facebook posts. Across these studies, I argue that user generated data is a source of collateral information that can augment clinical practice and potentially guide interventions.