Natural Sciences

What’s in a Tweet?

NIH funds two scientists to probe mental health via social media
twitter symbol with "Je Suis Charlie" written inside

After a gunman opened fire in a historic black church in Charleston, South Carolina, last year, killing nine people, thousands of tweets chronicled the tragedy and captured debate about it.

Within four days of the 2014 shooting death of Michael Brown by a police officer in Missouri, there were reportedly more than six million tweets with the hashtag for the city involved, #ferguson.

And within two days of the 2015 terrorist attack that took 12 lives at Parisian weekly Charlie Hebdo, the hashtag #JeSuisCharlie—“I am Charlie”—had been used 216,000 times, ABC News said.

When tragedy strikes, huge groups of people post hundreds of thousands of messages to Twitter and Facebook to vent, to engage, to act. That makes these digital sounding boards fertile ground for experts who want to study mental health.

After a trauma such as a mass shooting, the messages that someone posts to Twitter or Facebook often say something about them: their personality, what emotions they’re experiencing, whether (or how) they’re coping—perhaps even what actions they’re contemplating.

Embedded in those messages are important indicators of mental health—that is, the status of one’s emotional, psychological or social well-being. Experts can study those messages to advance understanding of mental health issues, not just for an individual but—given the reach of a platform such as Twitter—for an entire community or region.

The challenge is to determine how to interpret the records of behavior available on Twitter. Now two UO scientists will do just that.

The National Institutes of Health is funding a project by psychology professor Sanjay Srivastava and Reza Rejaie, a professor in computer and information science, to develop an approach for large-scale mental health research on Twitter. The two intend to use those ubiquitous, 140-character tweets to draw inferences about users’ personalities, emotions and potential clinical symptoms.

It might not be immediately obvious that analyzing tweets could improve public health, but the scientists believe there is enormous potential.

Srivastava studies how personality affects and is affected by the social environment. He said outcomes of the project could include information about how mental health indicators vary over time and between communities and regions; how major events such as disasters and mass shootings affect community mental health; and how social media might be used for individual-level screening and diagnosis.

“There are mental health variables that are common experience—depression, anxiety, post-traumatic stress,” Srivastava said. “We’re interested in individuals but also, at a more collective level, communities. What kind of difficulties are they having, and can we use that to do better health research and policies?”

Srivastava and Rejaie have developed a three-step approach. First, they’ll capture data from a large, representative sample of US Twitter users to determine what kinds of questions can be addressed.

Next, they’ll recruit a group of Twitter users to complete standardized assessments of personality, emotion and clinical symptoms. That will enable them to determine which online behaviors are reliable indicators of one’s particular psychological makeup.

Finally, they will monitor millions of Twitter accounts around the country for one year to study how one specific kind of community trauma—mass shootings—affects personality, emotion and mental health, both for individuals and large groups of people.

Rejaie, an expert with computer networks, has developed a technique to capture a representative sample of Twitter accounts from across the US, and collect their tweets. He will then run checks to ensure that the collected tweets include relevant messages—and aren’t simply links to cat videos.

Next, he’ll look for patterns within the messages, assigning values to specific words. Lastly, he’ll translate the results into information that Srivastava can use to make observations.

“This is a big shift from traditional social science, where a psychologist goes to an individual and asks them specific, individual questions,” Rejaie said. “This data is not meant for this purpose. You don’t have the opportunity to ask for what you want. You try to answer questions from what you have collected.”

—Matt Cooper

Photo credit: Dominique Dubuc