The Children’s Healthcare of Atlanta Pediatric Technology Center at Georgia Tech will evaluate an opt-in teen social media surveillance tool trained to monitor and detect mental health decline.

 

“The goal would be for smartphones to be agents of surveillance and safety just as they can be agents of adverse experience for young people,” said John Constantino, MD, Chief, Behavioral and Mental Health, Children’s. “Currently the parental controls for smart phones are not adequately validated for efficacy. This represents a novel approach to precision monitoring in real time.”  

 

Dr. Constantino will recruit teens ages 13-17 to test associations between social media activity and mental health status using an artificial intelligence (AI) tool. The tool was developed by Munmun De Choudhury, PhD, Professor, Interactive Computing, Georgia Tech, who has extensive expertise in assessing mental health from online interactions. Her lab has previously demonstrated meaningful signals unique to each person that can help predict future adverse outcomes, such as behavioral and linguistic patterns and if they are receiving negative feedback or support. 

 

Patients seen at Children’s emergency departments for severe mental health concerns who are now receiving follow-up outpatient care at Zalik Behavioral and Mental Health Center will be able to opt in to the study. The researchers will collect electronic health record (EHR) data from 200 participants who will also donate past social media usage data. Drs. Constantino and De Choudhury will examine past ED visit timing alongside past social media behavior to fine tune the tool, a model Dr. De Choudhury trained on social media and propensity to influence suicide, to see if it can make accurate predictions. 

 

Once the researchers know it can make accurate predictions, it will be designed to monitor users’ social media activity in real time and send alerts if certain behaviors are noticed that fall under one of three levels of urgency. The first level alert occurs when the tool notices a teen’s social identity deteriorating but not to the level of an emergency—in this instance it would send a direct message to a user encouraging them to reframe social media experience and redirect or pause its use. The second level alert occurs when the tool notices a behavior which needs caregiver intervention, in which case a direct message would be sent to the user’s parents. The third and highest level of alert triggers 911 because the tool has identified an imminent risk of harm to the smartphone user or others in his/her social network.   

 

“It’s very clear that social media experience can be toxic and at times potentially fatal,” said Dr. Constantino. “We’re hopeful we can harness technology to monitor social media use among youth whose own intention is to keep safe, toward the development of a scalable mechanism for both surveillance and support, such that if social media activity is predicting a major adverse life event, we can intervene effectively and on time.”