University of Wisconsin-Madison Center for Healthy Minds to study machine learning predictions of well-being

High rates of depression, anxiety, loneliness and lack of purpose are taking a toll on Americans’ mental and physical health, with wide-ranging negative consequences. Unfortunately, many of those most in need of mental health services are unable to access them through traditional means.

Researchers at the University of Wisconsin–Madison’s Center for Healthy Minds are working on one possible solution that leverages artificial intelligence to deliver personalized well-being content through a person’s mobile device when they need it most.

“We hope to discover how to offer supportive well-being practices on mobile phones that are sensitive to what a person is feeling and easy to engage with in the moment,” says Christy Wilson-Mendenhall, co-investigator on the new, two-year project, which is set to launch this year.

Supported by a $3 million-plus grant from the Defense Advanced Research Projects Agency (DARPA), the project aims to discover and develop technologies that in the future may inform the delivery of algorithm-based micro-supports — short well-being practices like mindful breathing that may help in the moment — through a version of the Healthy Minds program, currently used in the Healthy Minds Program smartphone app.

The multi-phase project will begin with a research study with human participants to predict their momentary emotional states, especially when they are feeling distracted, lonely, self-critical or unfulfilled, times when they may benefit from micro-support.

This initial step will help inform the development of machine learning tools and algorithms for the new platform. The researchers intend to develop their micro-support framework and content through testing with a diverse group of participants to examine whether users find the micro-supports relevant, timely and potentially helpful. The end goal is a platform that can deliver micro-supports based on an algorithmic trigger.

“In many ways, we are at the very beginning of finding how best to incorporate mobile technology into our daily lives in ways that support, rather than reduce, our well-being”

Simon Goldberg

In practice, the app will gather information from consenting participants through in-the-moment “experience sampling.” This sampling will entail brief, user-created videos in response to prompts like “I am feeling…,” as well as passive mobile data like geolocation and activity patterns.

Researchers will analyze the data to assess users’ well-being in terms of the four pillars of the Healthy Minds well-being framework: awareness, connection, insight and purpose. Previous research has linked these pillars to trainable skills associated with well-being. This analysis will inform delivery of very short, personalized “in-the-moment” well-being practices to keep users engaged in cultivating well-being behaviors throughout the day. These micro-supports will be easy and minimally disruptive, allowing users to be fully present with their day-to-day activities.

This novel research furthers the development of an evidence-based, scalable option in personalized mobile well-being interventions, which to date have used a one-size-fits-all approach with inaccessible costs and time commitments, resulting in early drop-off by users.

“There are rich streams of data that can be acquired through low-effort activities that individuals are accustomed to doing on mobile phones like creating a short video, and also completely passively from their phones, with their consent,” says Richard J. Davidson, founder of Center for Healthy Minds. “This type of data may be an informative context for tailoring delivery of specific micro-supports that may increase well-being.”

Davidson believes micro-supports delivered at the right time and place may have an outsized impact on well-being.

“Imagine how an anxious student receiving a prompt for a 30-second mindful moment right before a final exam might perform better, or how a meeting might unfold after a leader receives an appreciation micro-intervention right beforehand,” says Davidson. “There are endless opportunities that can be tested in real-world settings.”

The study will position researchers to eventually conduct a randomized control trial to test the efficacy of micro-supports triggered via machine learning tools.

“In many ways, we are at the very beginning of finding how best to incorporate mobile technology into our daily lives in ways that support, rather than reduce, our well-being,” says Simon Goldberg, an assistant professor of counseling psychology at UW–Madison and core faculty at CHM. “This project and the kinds of micro-supports that might be triggered based on what we learn can be part of developing technological tools that promote a healthy quality of mind.”

Future aims include continuing research and eventually releasing a platform for widespread public use through a range of web, mobile and wearable devices.