• by Erik Sass , Staff Writer @eriksass1, June 4, 2014

Reports of online bullying have soared over the last year, according to a new report from McAfee, titled “2014 Teens and the Screen study: Exploring Online Privacy, Social Networking, and Cyberbullying,” which states that 87% of teens and preteens polled in this year’s survey said they have witnessed cyberbullying, up from 27% last year.

Among respondents who reported being bullied online, 72% said it was due to their appearance, while 26% said it was due to race or religion, and 22% said it was due to their sexuality. Among youths who witnessed cyberbullying, 53% said the victims became defensive or angry in response, and 47% said the victims deleted their social media accounts to escape the abuse. Around a quarter (24%) of respondents said they would not know what to do if they were harassed or bullied online.

Perhaps even more alarming, social media is also playing a role in prompting “real world” confrontations, McAfee found, with 50% of the young people surveyed reporting being involved in a face-to-face argument because of something posted on social media (up from 33% last year) and 4% reporting getting in a physical altercation because of an argument that started online.

Last week I wrote about a report that nearly 2,000 children, some as young as nine years old, have been investigated by police in the U.K. over the last three years for potentially criminal activity on social media, including abuse, threatening messages, and online bullying. Previously a separate survey of British teens found that 21% had posted negative comments, while 26% said they had hijacked someone else’s account to post damaging content, and 43% said they had exchanged messages with strangers.

On the positive side, researchers with MIT Media Lab’s Software Agents Group claim to have developed an algorithm that can detect bullying, based on linguistic patterns. The program is designed to work in conjunction with social-media sites, as a kind of preventive measure. It scans the text of posts and when it detects language that may indicate bullying, it asks the user: “Do you really want to say this?”