Andrew T. Campbell

Mental illness touches almost every family these days. It negatively affects mood, behavior and functioning, and sadly still remains stigmatized even in these more enlightened times. Symptoms vary depending on the type of illness. Take depression on college campuses for example. First year students entering college can quickly feel stressed, anxious, isolated and overwhelmed, making them more susceptible to depressive episodes, dropping out of college, even suicide. People living with serious mental illness such as schizophrenia can suffer severe hardship including homelessness, victimization and incarceration. The current approaches have failed.

Over the last decade, we have seen the rise of mobile sensing capable of assessing, tracking and predicting trajectories of mental health using mobile phones and wearables. However, most reported results are limited and come from small scale studies. If this nascent “mental health signal” proves to generalize across different populations and types of illness (e.g., anxiety, bipolar) it will have profound and long lasting impact on the health and wellbeing of over 450 million people living with mental health illness on a daily basis worldwide.

In my talk, I will first discuss results from some recent projects on depression and schizophrenia sensing. I will make the case that future mental sensing technology will be mobile first and driven by AI. It will combine personalized sensing and intervention at population scale, and lead to new breakthroughs potentially directly connecting human behaviors passively sensed and inferred from mobiles to brain functioning. I’ll argue that a decade or so from now mental health sensing will be cheap, robust, mobile, ubiquitous and always-on. Ultimately, it will replace traditional episodic clinic-based assessments, which has left millions of people behind because of the lack of access to services. To democratize mental health and make mental health sensing globally accessible to everyone with a phone much more research is needed. Importantly, a business case still needs to be made. My talk is a call to arms: we roughly know how to do this technologically - so let’s get it done, why wait?