Inferring Human Behavior using Mobile and Wearable Devices

Abstract

Mobile, wearable, and ambient sensing is making possible the inference of activities and behavioral patterns of individuals and populations. This data-driven approach to discovery can help determine how these behaviors affect our health, as well as to assist in interventions aimed at promoting the adoption of healthier habits. I describe recent progress in this area, as well as some of the open issues that need to be addressed, and which provide opportunities for future research. These issues include the development and deployment of sensing platforms; identifying activities and behaviors that are relevant to healthcare and that can be inferred with sufficient precision using existing sensors; creating and curating large datasets and associated analysis methods from which strong evidence can be derived; and envisioning novel application scenarios that make use of the behaviors monitored. Several examples are provided to illustrate recent advances and open issues, including the use of pervasive videogames to assess frailty; using wearable devices to detect anxiety in caregivers of people who suffer from dementia; and using crowdsourcing to monitor and modify eating behaviors. Finally, I propose a new frontier in mobile sensing for healthcare, namely, inferring how individuals perceive themselves with respect to others in order to change these perceptions and improve their wellbeing.