Passive data collection
How ginger.io innovated mental health tracking via passive data collection
Many years ago, I met Anmol Madan, soon after he got his PhD in machine learning from MIT.
Anmol was busy creating the company ginger.io (later acquired by Headspace) to use passive data collection for patients with mental health issues.
Anmol’s insight was that manual data collection for wellness was a horrible idea. There are over 100 mood tracker apps on the iOS AppStore, but 95% of them are abandoned after 10 days.
Only some of our OCD friends are able to consistently give input to a manual process each day. The vast majority of us fail at this simple task.
What Anmol did instead was to create software (initially for Android OS) that would sit in the background on a user’s mobile phone and observe patterns that were created not manually but automatically. Stuff like this:
Use the accelerometers and GPS to detect how long it has been since the phone has moved? If a patient usually gets out of bed at 8am to start their day, but then one day the phone has not moved until 2pm, that’s a possible warning sign.
Observe the number of calls/emails/texts made and to whom, and to look for unusual patterns. If someone who usually sends about 5-10 messages an hour suddenly starts sending about 50 an hour, that could be a sign that they are entering into a manic state. On the other hand, if that same person starts messaging only 1-2 times an hour, possibly they could be entering a depressive state.
At BVS, we are building passive data collection – via Apple HealthKit and Google Health Connect – to observe data that the user allows, from devices such as FitBit, Whoop, Apple Watch, Aura Ring, or the phone itself. We’ll use signals from these apps in our AI apps to help us assess how our patients are doing.
Thanks to Anmol and his ginger.io team for waking us to this strategy, and for all of their years of work to prove out the best ways to do passive data collection for mental health.