Frisco Students Develop Self-Harm Prevention AI

Photo: mooremedia | Shutterstock

After the COVID-19 pandemic put school attendance on hold, Centennial High School seniors Lavik Jain and Shivansh Nikhra from Frisco developed an AI that detects suicidal and self-harm posts on social media. Now the Denton County MHMR Center adopted the system to contact and guide people who expressed self-harm thoughts online. 

Faced with more time on their hands when almost all social interactions were online, the two high schoolers started exploring machine learning and artificial intelligence. Soon they discovered many people were more comfortable opening up and sharing personal experiences and feelings on social media than in person. 

“We saw people expressing themselves online much more,” Jain told North Texas Daily. “Social media is a platform where, whether it be because of the anonymity or lack of face-to-face interaction, people express themselves more.”

This was the inspiration for Project Suicide Watch, a machine learning-based model that in addition to detecting suicidal social media users, helps them find the help they need by connecting them to healthcare and public service organizations. 

According to a Linkedin post by the Denton County MHMR Center, 2021 saw a total of 98 suicides and another 51 by September 2022. In November, the center adopted a trial run. 

“They did it in over a week’s time just to see how many individuals would have tweeted suicidal statements and only one individual came back, which was good,” MHMR chief operations officer Brittany Waymack told  North Texas Daily. “So, we knew it wasn’t gonna be hundreds of people coming back.”

The system first reads all the incoming social media posts within a certain region every five minutes, once the model processes all the posts, it identifies those that are deemed suicidal and lastly, moderators connect these social media users with help and are added to a spreadsheet for each city. Cities receive this spreadsheet every week. 

“You can think of machine learning as a black box: when fed some input, it is expected to spit out an output,” reads the Project Suicide Watch website. “This black box has been trained with 600,000+ example social media posts… The model does not look for specific words or phrases; rather it detects an underlying suicidal tone.”

The first organization to start using the system was the Wylie police department. This partnership helped Jain and Nikhra get in contact with several organizations. Currently, the team has partnered with Texas Health and Human Services, Mental Health America of Dutchess County, Irving and Frisco police departments and Denton County MHMR Center. 


In case you missed it, here's Local Profile's report on Texas' new Bill to ban social media for underage children.

Return to Local Profile