#1 social network in Russia fights bullying with AI

The Man

Former Staff
Jul 2011
48,507
36,334
Toronto
VKontakte really wants you to watch what you’re saying.

Russia’s largest social network conducted a one-day test of a tool that urges users to think twice before sending insulting comments, the company said in a press release Monday.

An algorithm detected abuse and recommended the sender “not to waste time on aggression and avoid the offensive remark.”

“Write something kind. Surprise them!” said a sample pop-up message over an unsent text calling the author of a post “dumb trash.”

VKontakte said it picked Russia’s National Unity Day public holiday to test the new feature, then gauge how effectively it “reduces the amount of negativity” online. On all other days, the social network reminds users they can still report bullying, harassment and other offensive content through a built-in button.

“This will help protect adolescents who are more likely to experience bullying, and prevent harassment based on ethnicity,” the social network said.

Users reported being able to dodge the algorithm, and others suggested that VKontakte’s experiment had backfired.

“I cursed out and swore at a bunch of my friends as an experiment, and VKontakte didn’t stop me,” Russia’s Business FM radio station quoted an unnamed user as writing.

VKontakte, run by Russia’s internet giant Mail.ru Group, did not say when or whether it plans to roll out the decency function in full.

As you see, results are questionable, thus far. Hopefully it improves, eventually...
 

Djinn

Council Hall
Dec 2007
52,418
39,535
Pennsylvania, USA
Years ago, I was surprised when I received an automated e-mail from my employer's Outlook server. It said that one of my recent e-mails had been flagged as "abusive" and would be reviewed. I had several friends in IT, and I asked them about it. They said that lots of e-mails get flagged. They're not normally reviewed by anyone; they're stored for 30 days, and if no managers or execs request it, they're deleted.

But I was puzzled... I didn't see anything in my e-mail that would have caused it to be flagged for abusive language. It was an e-mail sent to multiple employees in which I was coordinating a paintball game later that month. The IT guy found it ... and pointed to a sentence where I called Ray (a co-worker) a "pain in the ass." I told him that I routinely called people a "pain in the ass." I told him that as a time-saving measure, I was planning on changing my autosignature to "Thank you; you're a pain in the ass."

The difference was that this e-mail wasn't just going to internal employees. Some of the people involved in the painball game were former employees with external e-mail addresses, and because the e-mail contained the language it did, PLUS the fact that it was going outside the company, that's what caused it to be flagged. However, the IT guy did agree with me that Ray was absolutely a pain in the ass, and I should go out of my way to shoot him on the paintball field.