The way that we use our technology is influenced by our psychological state and our health. How fast we type, how steady we hold our smart phone, and how often we check our phone can be indicators of mood and anxiety. The pattern of battery charging tells us about circadian rhythm and sleep disruption. Every smart phone contains powerful digital sensors that can be used to monitor these “phone behaviors”, which can be viewed as phenotypes of our psychological and physiological state. By mining the sensor data generated by smart phones it is possible to passively detect changes in a person’s behavior, and algorithms can then be used to help predict specific psychological constructs and potentially deliver personalized behavioral interventions. This concept was first explored by the MIT Media Lab in 2011 as part of early work on Android (funf.org) and a subsequent start-up company (Behavio), and was refined and formally named “Digital Phenotyping” by Jukka Onnela in ~2015.
There are many people who cannot use, or cannot afford, wearable sensors. In these circumstances, the existing sensors embedded in smart phones can be used to passively collect data from users, after consent and permission is given. In addition to these sensor data, other digital phone information, such as call/SMS activity and GPS location can also be used to explore other dimensions of user behavior and mobility.
Our group has developed it own digital phenotyping platform and algorithms, with particular emphasis on preserving privacy and anonymity to enable mobile health researchers to make use of smart phone data. Applications include substance abuse and bipolar disorder. In addition to smart phones, digital phenotyping data are also available from other devices in our environment, which can range from light switches, televisions, refrigerators, and other digital appliances. Smart home digital phenotyping is now developing into its own field of research with applications to patient monitoring, senior care and aging.