Researchers leverage Twitter to identify cyberbullying

One of the biggest problems that schools deal with today is online bullying. It’s an interesting issue because—for the most part—it happens outside of the school building. Still, when a tragic event occurs as the result of bullying, faculty and administrators are questioned on what steps they took to intervene.

The problem is further complicated by the fact that “snitching”—or reporting bullying to a teacher or other adult—is considered taboo and will often escalate the situation. Even when students do report bullying to adults, it is often brushed off and not taken seriously.

According to Amy Bellmore, a educational psychology professor at the University of Wisconsin–Madison, "Kids are pretty savvy about keeping bullying outside of adult supervision, and bullying victims are very reluctant to tell adults about it happening to them for a host of reasons," She added, "They don't want to look like a tattletale, or they think an adult might not do anything about it."

Despite the stigma around confiding in adults or alerting them to the fact that a peer is being bullied, there are ways that schools and parents can uncover bullying and other forms of abuse: social media. Researchers at the University of Wisconsin—Madison have crunched the numbers and developed an algorithm churns through data pulled from Twitter.

The program capitalizes on Twitter’s public API to scan the over 250 million tweets published every day. An initial pilot uncovered over 15,000 tweets that involved bullying. Depending on the content of the tweet, the researchers were able to determine additional information, such as whether or not the tweet originated from an observer, the victim, or the actual bully.

Traditionally, most studies on bullying in schools draw from much larger sample sizes—typically, 20-30 students—and rely on student testimonies and self-reports, which suffer from the “snitch” bias. Additionally, they are one-time snapshots. It is unlikely that a researcher will be able to access a given set of students over a period of time. 15,000 tweets may not seem like much next to the sheer volume of tweets posted daily, but it provides a much richer set of data than researchers have had access to in the past. Researchers can also tap into social networks to analyze reports from a given user over a period of time and potentially compare and contrast different perspectives on the same event by leveraging follower relationships on the social network.

The algorithm capitalizes on a technique known as machine learning, which is founded on the idea of leveraging computers to develop algorithms that can teach themselves to become more accurate as they encounter more data. Google uses this technique to deliver more accurate search results.

"The computer gets a set about bullying and a set definitely not about bullying," says Jerry Zhu, a computer science professor at the university who studies machine learning. "In machine learning, the algorithm reads each tweet as a short text document, and it goes about analyzing the word usage to find the important words that mark bullying events."

Once trained, the computer algorithm can process more data than an researcher alone and become smarter as it goes along. The result is that schools and parents may have more insight into their child’s safety and well being than ever before and tragic events such as teenage suicides that have resulted from bullying will be more likely to be intercepted in the future.

Photo credit: Bartmoni on Wikimedia Commons.