Tracking Mental Health Through Twitter Analysis

1-29-2014 10-14-18 PM

Researchers at the Center for Statistics and the Social Sciences at the University of Washington are scouring Twitter for signs of depression in individual users. Following in the footsteps of many other public health scientists, these researchers hope to repurposed otherwise useless social media posts by using them to uncover patterns in the way that we present ourselves externally across a variety of different mental states.

Tyler McCormick, a researcher at the University of Washington, explains “Our attitude is that Twitter is the largest observational study of human behavior we’ve ever known, and we’re working very hard to take advantage of it.” McCormick will lead a group of researcher that hope to take a basic algorithm developed by Microsoft researchers and improve its ability to identify depression in users.

The Microsoft algorithm was a first attempt at uncovering a picture of mental health through twitter analysis. That project looked at up to a year’s worth of twitter history on each user, evaluating variance in overall volume of tweets, what times individuals tweeted, and how frequently they engaged other users directly. The algorithm also scanned for keywords that researchers found a had a strong association with depression. The final result was a tool that could accurately predict depression with a 70 percent accuracy, but missed depressed users as much as 30 percent of the time, and misidentified healthy users as depressed 10 percent of the time.

Now, researchers hope to refine the algorithm and improve the performance results. His team will delve into the erroneously flagged users to uncover what data is misleading the algorithm. They will also look for new patterns within depressed populations that will help it increase its overall accuracy.

A similar study currently underway at the University of California, San Diego is trying to extract mental health information at a population level. Researcher’s there are looking at ways of more effectively using limited mental health resources in a community by identifying the most at-risk groups in a more real time manner.

If successful, McCormick imagines a scenario where social media posts could be used to passively monitor the mental health of at risk populations, such as college undergraduates and veterans returning home from deployments. The tool would might be implemented in such a way that it could trigger preventative interventions in users that might have otherwise gone untreated.


Enjoy HIStalk Connect? Sign up for update alerts, or follow us at @HIStalkConnect.

↑ Back to top

Founding Sponsors

Platinum Sponsors