Detecting Cyber Bullying: But Can it Be Stopped?

Mar 2, 2017

Credit Marcie Casas / Creative Commons

In the Internet age, some every day problems have made an easy leap to the virtual world.  Take bullying.  It can be bad enough in person, but online, it’s growing fast and may be causing even more damage.

“You're ugly. Your mom hates you. You’re no good. You’re not a man, you don’t have any friends…."

The words of bullies everywhere are remarkably similar.  Personal attacks designed only to be hurtful.  They cause real damage. And they’re not confined to children in the schoolyard. In fact, they’re making up a larger percentage than ever before of communication among adults online.  

“I see it a lot with people and politics.  If someone gets into a political argument and just does not stop making their point over and over again and not listening to the counterpoint,  that can become harassment very easily."

Bert Huang teaches computer science at Virginia Tech. He says the advent of social media fueled by computational power is expanding the scope of bullying.  And while its still occurs more often among children and adolescents, according to a 2016 study by the Data and Society Research Institute, nearly half of all adults have been the victim of online bullying and nearly three quarters have witnessed it.  

“Bullying tends to take the form of repeated attacks.  And the attacks can be different for different individuals, but a lot of times it’s that an attacker identifies something that is injuring the victim and will then repeat that sort of attack.”

Other forms of cyber bulling include, blocking access to online platforms or invasions of privacy. A slightly more sophisticated, but not a more grown up version of the original. 

“On the Internet, says Huang, if someone posts something that’s hurtful to you, a lot of times it’s just there forever and you can’t delete it. In traditional bullying you can get out of the space that the bully shares with you. You can physically remove yourself from the situation. This is a qualitative difference in that there is no real safe space that you can get to when it’s happening over the Internet.”

Virginia Tech Assistant Professor of Computer Science Bert Huang uses machine learning to detect cyberbullying.
Credit Virginia Tech Department of Computer Science

So Huang launched a project that harnesses the power of computers to join the fight. It uses machine learning to scan open social media sites like Twitter and Instagram, to flag word patterns that connote bullying attacks.  But while computers are good at the heavy lifting of mining the language on the Internet….

“When we talk about getting machines to do cyber bullying detection and many tasks like this that involve social contexts, we’re mostly talking about getting them to do a rough job a human can clean up later on.  That’s because human relationships are really subtle and computers are not that good at doing that yet, but what they are good at is doing lots of them really fast.”

It’s an example of how computer science is merging with social science to explore new problems. Like this one; once you identify the signs of bullying, what exactly do you do about it?  The whole project raises new ethical questions to which there are no easy answers.

“One of the problems with the current state of the Internet and this problem of toxicity and cyber bullying is that the only really good solution is to just shut it off.  But that’s really unfortunate because of all the great things that social media gives us, the ability to communicate and connect with individuals across the world; all these people who are forced to quit social media lose all that and it doesn’t seem fair because essentially whoever is doing the cyber bullying has taken that away from them."

Huang says at this point, the incentive structure doesn’t’ seem to be there for the social media companies to do much to stop the nonconstructive criticisms that fly around cyberspace; they make money when more eyeballs view explosive fights online.  He thinks it will ultimately have to be a third party solution.  So he hopes to enlist fellow researchers to join the project and take it to the next step.  And it makes you wonder, could machines help crack the code of this age-old trait so pervasive in the human condition, we call bullying.  Or will they simply make it faster and easier for it to continue?