Clemson University app could help fight cyberbullying
CLEMSON — Clemson University researchers are building what could be the first app to search photos and videos for signs of cyberbullying as experts grow concerned that defenses are falling short in the face of changing technology.
A research team representing three of the university’s five colleges announced Wednesday it has landed a $240,000 grant from the National Science Foundation to find new defenses against cyberbullying.
Over the next two years, the team will develop an app they are calling VC_Defender that would scan pictures and video on social media. It would be similar to facial recognition software, except that the app would look for cyberbullying images instead of faces.
The app would use the images and other information, such as social media posts and comments, to automatically figure out whether an adolescent is at risk for cyberbullying. It would assess attack severity and offer intervention strategies.
An alert would be sent to parents or teachers when it appears someone is being cyberbullied. The person who is doing the cyberbullying could also receive a message explaining that his or her behavior is hurtful.
Cyberbullying defenses have fallen short in the age of cell-phone cameras and instant posts to social media, researchers said.
While use of social media has boomed in the past decade, automatic cyberbullying detection techniques remain in the nascent stages, they said. Studies have mostly focused on cyberbullies’ written text and have largely overlooked the use of visual media, researchers said.
Many schools have programs to deal with traditional bullying, but cyberbullying programs have lagged behind, even as a growing number of schools provide iPads and other mobile technology for school use, they said.
Eighty-six percent of U.S. adolescents used the Internet in 2012, an increase of 67 percent from 2000, according to a 2013 report from the Center for the Digital Future. Each spent an average of 20 hours per week online, up from nine hours in 2000.
Robin Kowalski, a professor of psychology, said that cyberbullying is a serious problem that differs in several ways from traditional bullying.
It used to be that bullying mostly happened at school, but social media is always available, giving victims little respite from their torment, she said. And while the bullied once had no question about who was doing the bullying, cyberbullying makes it easier to launch attacks anonymously, Kowalski said.
Some adolescents won’t tell their parents about cyberbullying out of fear they won’t understand, she said. Some victims are afraid that the phone or tablet– often a social lifeline– will be taken away, Kowalski said.
“Cyberbullying can have a deeper and longer lasting effect than traditional bullying,” she said. “It can result in emotional, psychological and physical problems and poor academic performance.
“In the most catastrophic cases, it leads to suicide. It’s urgent that we develop defenses against cyberbullying attacks.”
Hongxin Hu, an assistant professor in the School of Computing, is the principal investigator on the grant. Co-principal investigators are Kowalski, Luo and Joseph Mazer, who is associate professor and associate chair of the Department of Communication Studies and director of the College of Architecture, Arts and Humanities’ Social Media Listening Center.
Researchers said that the biggest challenge they face in developing the app is creating a “classifier” to distinguish which images constitute cyberbullying. After all, a picture or video can have different meanings depending on the context.
The classifier, which will be part of the app’s software, will build profiles of social media users based on publicly available information. Factors could include everything from the music and books users like to the posts and comments on their accounts.
“No existing system does what we are proposing,” Hu said. “Once we have the solution, we hope it will be adopted by popular social network providers.”
The first step in the research will be to collect and analyze data that will help the team better identify cyberbullies and their victims.
The team will use the Social Media Listening Center to study actual cyberbullying attacks and assess trends.
Through the center, researchers will be able to measure and engage in 650 million sources of social media conversations, capturing publicly available data from Facebook, Twitter, YouTube, blogs, online communities and mainstream news outlets. The center has a 152-inch video wall and is powered by Salesforce Marketing Cloud’s Radian6 software.
“The Social Media Listening Center will allow us to easily identify, categorize and analyze cyberbullying attacks,” Mazer said. “We’ll use the information to create an online intervention system that identifies and addresses these harmful behaviors.”
Funding comes from the National Science Foundation’s Early Concept Grant for Exploratory Research (EAGER).
Deans from the three colleges issued a statement congratulating the team on landing the grant. The statement came from Anand Gramopadhye, dean of the College of Engineering and Science; Robert McCormick, interim dean of the College of Business and Behavioral Science; and Richard E. Goodstein, dean of the College of Architecture, Arts and Humanities.
“It’s exciting to see Clemson researchers coming together to address an urgent problem affecting adolescents around the world,” they said. “This is an excellent example of interdisciplinary scholarship, and we look forward to supporting the team.”
Kowalski, who is one of the authors of www.cyberbullyhelp.com, said the research could eventually expand beyond adolescents to include adults.
“Cyberbullying is alive and well in the workplace,” she said. “It’s great to start with kids and take it from there.”
This material is based upon work supported by the National Science Foundation under Grant No. 1537924 . Any opinions, findings and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the view of the National Science Foundation.