The rise of artificial intelligence has created new opportunities for computer coding, scientific research and a host of other productive uses, but it has also made it alarmingly easy to create and spread content that is potentially harmful to young people.
An interdisciplinary team led by Clemson University is responding by launching new research that includes better understanding how generative AI makes decisions and developing new safeguards.
The team is beginning its project as generative AI platforms become increasingly adept at generating photos and videos, some that are difficult to distinguish from reality.
Of particular concern to the research team is sexually explicit or abusive material, toxic language and other not-safe-for-work content that could affect young people’s mental health. Such content has been linked to anxiety, cyberbullying, isolation and eating disorders, researchers said.
The project’s principal investigator is Linke Guo, an associate professor in the Holcombe Department of Electrical and Computer Engineering.
He said the idea was partly inspired by his daughter and his concerns about the kinds of harmful AI-generated content young people can encounter online.
“One thing I care about is the mental health of kids,” Guo said. “I worry that if kids can easily access this content prematurely for their age, then it will create bad consequences when they enter adulthood. I want them to grow up in an environment that is friendly, clean and free of this kind of harmful content.”
The co-principal investigators include three from Clemson: Long Cheng, associate professor in the School of Computing; Heather Hensman Kettrey, associate professor of sociology; and Matthew Costello, associate professor of sociology. Also collaborating is Ming Li, associate professor of computer science and engineering at the University of Texas at Arlington.
The team will study how generative AI interprets prompts and produces unsafe content, then build safeguards that can detect hidden intent, remove unsafe elements from images and check generated images and videos.
They will also survey and interview young adults and parents to see whether those safeguards reduce harmful exposure and change how people use and respond to AI.
The interdisciplinary project brings together technical expertise in artificial intelligence and computer science with research in sociology and criminology so the team can evaluate both how well the safeguards work technically and how young adults experience and respond to them.
The four-year project is supported by $1.2 million from the National Science Foundation through its Secure & Trustworthy Cyberspace program.
Zhuoyang Wang, a Ph.D. student in electrical and computer engineering, will bring a first-hand view of AI to the project. She said she has been using AI since 2019 when she was 17 years old and has seen how quickly generative AI has become part of young people’s daily lives.
“Sometimes it is helpful, and sometimes it is just for fun, like creating funny pictures of friends,” Wang said. “But some may raise some concerns about whether it is safe and appropriate.”
Hai Xiao, chair of the Holcombe Department of Electrical and Computer Engineering, said the research addresses an increasingly important challenge as generative AI becomes more powerful and more widely used, particularly among young people.
“This is important work because we need to make sure advances in artificial intelligence are accompanied by strong safeguards,” he said. “Linke is well positioned to lead this effort because of his expertise in AI security and his ability to bring together researchers from several disciplines to address a complex problem.”
