AI tool aims to catch teen mental health crises using social media data - WRDW

ATLANTA, Ga. (WRDW/WAGT) - Georgia Tech and Children’s Healthcare of Atlanta (CHOA) are working together to develop an AI social media surveillance tool that can help identify when a young person is at risk of a harmful mental health event.

The tool is based on a decade and a half of research. Researchers discovered signals from the types of words used online, what teens are engaging with and the types of information people are exposed to and share.

“We hope we will be able to build a technology intervention that can provide the right kind of support and help to a young person at the right time,” Georgia Tech Professor Dr. Munmun De Choudhury said.

Researchers found that social media data can be incredibly valuable to determine risk.

“Their social media profiles are a very important part of their lives,” De Choudrury said. “If we are not looking there, we will be missing a very important piece of their life that might be very valuable to prevent bad things from happening.”

The data is being donated by CHOA patients between the ages of 13 and 17. Health data is already sensitive, but even more so for teens — and on top of that, this is data not typically associated with health. Researchers work with families to be transparent about how the data is being used.

“When we build that relationship of trust with the family, many families are willing to share this kind of data with the research team because they see the value of science that can result from this kind of donation,” De Choudhury said.

Researchers hope the tool will save lives. While only about 6% of visits to pediatric emergency rooms are for behavioral crises, about 45% of those are related to suicidal thoughts or actions.

“What we’re trying to do is make that device safer,” Dr. John Constantino, chief of behavioral and mental health at CHOA, said. “Not only to use these miracles of technology, which are smartphones, but also to make sure that we’re putting on the guardrails that are going to prevent them from going down these kinds of difficult roads that can lead to catastrophic consequences to them.”

The AI tool is not meant to replace existing parental control, but it could be another way for families to prevent tragic outcomes like suicides, researchers say. And it’s not just looking at overt signals.

“What AI also allows us to do is to combine otherwise what might seem like benign cues, but when you put them together in intelligent ways, you can begin to see a pattern that can be alarming,” De Choudhury said.

The AI tool sends three levels of messages, all of which are currently being reviewed by researchers. The first stays solely with the teenager, encouraging them to pause, rethink what is happening and get help if needed. If the situation needs to go right to an adult, the AI would determine who to alert, whether it be a parent, guardian or even the authorities.

The messages are being tested with teens to see whether what seems like a good message is actually helpful. But the doctors say the AI would only send one to three messages.

“It is not designed to be a therapist or a buddy or anything like that,” Constantino said. “Rather, it’s just an agent of opportunity for flagging for that youth themselves when the indications are that this is starting to hurt them.”

Copyright 2026 WRDW/WAGT. All rights reserved.