For most people, a selfie's just a selfie. But looked at the right way, tweets and online posts may be very revealing about the way people and communities think and feel.
Researchers are beginning to tap hundreds of thousands of tweets emerging from conflict zones to see what the trove of data can reveal about the mental health of people in distress.
Signatures that these social scientists have identified — commonalities in word choice, posting frequency, and other factors across hundreds of people — could help direct aid to the groups who need it the most, they say.
A community numbed by violence
A new analysis of two years of Twitter data from four cities in Mexico investigates how communities are reacting to the horrors of the drug war after years of bloodshed.The conflict between Mexican drug cartels and the government is marked by feuds between rival drug lords and clashes with law enforcement and vigilante groups.
On Twitter, researchers have found signs that civilian bystanders have become numbed to shocking images or reports of violence, even when that conflict seems to be on the rise.
The widespread violence and stifled media reports have led to citizens in conflict-affected cities relying on social networks to "inform and collectively grieve, critique, and express frustration about the violence in the streets," the report's authors wrote.
For the Mexican social media users, it's a much-needed way to express themselves, check on their friends, and spread news. For researchers looking for the effects of conflict on a community, it's a veritable data goldmine.
“People are not reacting as negatively as you would expect them to," Munmun De Choudhury, study co-author and a researcher at Georgia Tech, told NBC News.
The results of the study surprised Choudhury's co-author, Andres Monroy-Hernandez, a researcher at Microsoft who grew up in northern Mexico. After the Mexican military became involved in the crackdown on the cartels in 2006, Monroy-Hernandez often heard grisly stories about murders witnessed by his family and friends.
"We can make interventions to make sure that people receive the right kind of attention at the right time."
Because traditional media is threatened and often gagged, citizens use Twitter to report and spread news of fresh incidents. But as years ticked on, those conversations — both in person and on Twitter — took a turn.
The Microsoft researchers used algorithms to draw out indications of numbness to violence and other stress. Algorithms trawled the tweets for specific words and word groups, as well as counting the frequency of the messages and logging the time at which they were sent.
The researchers admit that the process and final results don't translate into simple tips on how to spot distress. Yet the study's findings reveal how Twitter can be a window to long-term effects of a population in crisis, Choudhury said.
"We can make interventions," she said, "to make sure that people receive the right kind of attention at the right time."
Filtering the stream of raw data
Choudhury is an emerging leader among a handful of researchers who want to tap the flood of social media data to gain insights into the minds of individuals as well as larger communities.
In two studies published in 2013, she demonstrated how algorithms analyzing the Facebook and Twitter activity of expectant moms could predict whether an individual is likely to suffer post-partum depression.
"When the population is subject to a long-term crisis, we lose that there's a longer term health concern that is happening – [that's] being ignored."
These algorithms filter the Twitter firehose, "picking out the flecks of gold from the silt," explained Glen Coppersmith, a Johns Hopkins researcher who also studies mental illness indicators on social media.
Choudhury's done a "fantastic job" recovering early signatures of some conditions, Coppersmith said, "but there's much work to be done." He is researching a host of other conditions, and at a conference in June will present early results from his team’s work spotting the signature of PTSD in the Twitter streams of people diagnosed with the condition.
Reaching back with help
While monitoring social media for signs of mental distress is still a nascent science, some aid organizations have already begun using Twitter as a way to reach out with mental health help after a disaster.
For days after the Boston Marathon bombings, American Red Cross volunteers sent messages of support on Twitter and Facebook to people who posted messages about being afraid or anxious.
"Emotional support is one of the most common ways we respond to people in emergencies on social media," Gloria Huang, Senior Officer, humanitarian social engagement told NBC News. In the days after the bombings, because damage from the two bombs was mostly contained, "a lot of the support we were providing [was] mental health support," she said.
"[They] logged on and just had conversations online about: it's okay to be scared by this, it's okay to be stressed by this," said Huang. The volunteers also circulated mental health and self-care tips.
Groups are also using similar techniques in Mexico. In the city of Monterrey, the Center for Citizen Integration reaches out on Twitter to coordinate aid and psychological support. The group monitors hashtags to find people who need assistance, and responds to people who tweet at the center with questions.
"When the population is subject to a long-term crisis," Choudhury said, "we lose that there's a longer term health concern that is happening – [that's] being ignored."