Ganging Up on Cyberbullying

This week, 14-year-old Jamey Rodemeyer from New York ended his life after being persistently bullied on a social networking website. Despite talking about being bullied and asking for help repeatedly, nobody did anything to comfort him. And so when nobody seemed to acknowledge or care about his pain, he tragically ended his life. Every time I’ve thought about this story, I’ve felt a range of emotions, from chills and sadness to outright anger and disgust that nothing was done to help Jamey.

The scourge of cyberbullying has increased alarmingly over the past couple of years. Defined as the use of communication technologies for “persistent and repeated” harassment of individuals, cyberbullying magnifies the effects of traditional bullying. The fact that many cases involving adolescent cyberbullying have ended with tragic outcomes like suicides underlines the grave nature of this menace. There has been national and international recognition of the hazards of cyberbullying at the highest echelons of power, from the White House to the British House of Commons.

Just like spam once threatened the viability of email, cyberbullying now casts a dark shadow over social networking websites. Most of the current work focused on cyberbullying involves surveys to gauge its prevalence and awareness campaigns led by parents and school administrators. While these approaches have their place, there has been very little use of technology to tackle this problem.

Last fall, after watching a segment on cyberbullying by CNN’s Anderson Cooper, I began to wonder if there were any efforts to use computational linguistics to help detect textual cyberbullying on social networking websites. My thoughts found immediate and strong support from a fellow graduate student, Birago Jones, and my Media Lab advisors, Dr. Henry Lieberman and Professor Rosalind Picard. We decided to pry further into this topic. We were perplexed to discover that next to nothing was being done in the field of computational linguistics to tackle this problem. The more deeply we thought about it, the clearer it became that a technical approach to tackling cyberbullying on social networks would have to be a combination of computational linguistics for effective detection and reflective user interaction techniques for changing user behavior.

Since that time, we have been lucky to work with a couple of social networking websites and with other collaborators from both within and beyond the Media Lab. This past March, we were invited to the White House to participate on a national summit on tackling bullying led by the President and First Lady, where a collaboration between the Media Lab and the social networking website Formspring was announced. Since that time, the White House has connected us to MTV, with whom we are now collaborating. This week, the Department of Education held a follow-up summit where Secretary of Education Arne Duncan announced that MTV, in partnership with the MIT Media Lab, is opening the corpus of teenage stories on subjects related to cyberbullying for the wider research community at the website athinline.org.

In a world where applied natural language processing and user interaction design tend to revolve around commercial applications, precious little is being done on tackling this very serious social problem. We are so excited here at the Media Lab to draw from varied disciplines such as psycholinguistics, affective computing, adolescent psychology, sociology and then apply it to exert the full power and weight of natural language processing, machine learning, and human-computer interaction to develop a practical and empathetic solution that social networking websites can actually implement. And we find this work to be deeply inspiring.

Karthik Dinakar is a second-year graduate student researching applied natural language processing and machine learning in the Software Agents group

Birago Jones is a second-year graduate student researching reflective user interaction in the Software Agents group

Henry Lieberman is a Principal Research Scientist and director of the Software Agents group

Rosalind Picard is Professor of Media Arts and Sciences and heads the Affective Computing group