Get news updates from WGBH
Mon January 13, 2014
MIT Algorithm Takes Aim At Social Media Cyberbullying
These days, students navigate social media sites as easily as the hallways.
And Max, Michael and Hyacinth — seventh graders at Weston Middle School, are among them. Each has an iPhone, and using mobile apps comes as second nature to them. They also know that social media can be a tool for bullies.
“Yeah, the big thing is, it’s anonymous,” Max said. “It’s sad, because if you could sort something out in person, the problem might get resolved. Online, no … ”
Hyacinth remarks on the impact of technology on social interaction.
“It has pretty much completely changed society in a really big way, I think cyber-bullying has definitely been one of the main things that occurs today,” Hyacinth said.
“Sometimes I catch myself before I post something,” Michael said. “Like, 'Will this be offensive to somebody?’ And there are many times that I don’t post something in case people disagree with it.”
Computing for Empathy
By nature most kids hit “send,” and until now, there’s been no way to track online bullying. But Karthik Dinakar, researcher at the Software Agents Group at the MIT Media Lab, thinks the lab has come up with a way.
“Bullying consists of so many different things," Dinakar said. "You can hack into somebody’s account, you can say really mean things, you can send really inappropriate pictures, but what I was most focused on was detecting things through communication when it happens through language.”
Dinakar developed an algorithm that matches what you write online to a database of commonly used words and slang, detecting whether it’s hostile or inappropriate. So for example, if I go to Facebook, type in a message and press “send” or “post,” the algorithm would analyze what I’m writing, and if it senses something nasty or mean, a little message would pop up, asking “Do you really want to say this?” Dinakar calls it “computing for empathy.”
“That’s one way in which it can be used," he said. "Parents can deploy the software on the phones of their kids.” Dinakar said.
You mean... have a conversation in person?
Right now the algorithm is being used at A Thin Line, an MTV website where teens share experiences. While it may not prevent bullying, it could help change behavior — which is something all of us could work on, according to Elizabeth Englander, psychologist and director of the Massachusetts Aggression Reduction Center at Bridgewater State University.
“Kids — and adults, too — have a tendency to increase their self-focus on online," Englander said. "So that means they really focus on how they’re appearing to others, and they stop focusing on how what they’re doing is going to impact other people.”
“The problem with that is that it may mean that they do something trying to look funny or clever or tough, and they’re really not thinking about how it impacts others," she said. "The MIT app is perfectly suited for that kind of situation because the purpose of it is really to remind the user that what they’re doing is going to have an impact.”
While Englander says there is a need for a self-flagging app, she says it’s only one solution to a multifaceted problem.
“One of the things that we need to teach our kids, is we need to teach them how to have a private conversation," she said. "How do you do it? Do you talk to the person face-to-face? Over the phone? But social media is not the place to have it.”
That’s a message that resonates with students at Weston Middle, where Englander’s anti-bullying credo is taught to students every school year.
Seventh graders Hyacinth, Michael and Max agree that the app would be useful.
“Yeah, because say something you say isn’t bad, but it’s thought that it’s bad,” said Max.
“Definitely”, chimed in Hyacinth. “I think it could give someone some awareness about like, it can be seen, other people can see it and it will definitely do some damage.”
"To have something pop up like that just refreshes your mind by telling you, are you sure?” Michael said.
A question we could all ask ourselves before hitting “send.”