Machine Learning Tackles Cyberbullying


Anyone who uses Twitter knows that it can be home to very hostile exchanges, some of which may even constitute cyberbullying. In a world where that kind of danger is a mere click away, what can parents do to keep their children safe without monitoring their every digital move? Analytics and machine learning are offering new options.

Bark CEO Brian Bason founded his company after having a couple of kids of his own and finding himself concerned about online safety. Bark offers a machine learning-backed app and service to help parents work together with their children to navigate the dangers of today's digital world. Bark launched in January 2016 after about half a year of machine learning training and testing.

In a phone interview, Bason told All Analytics that the development process involved extensive testing and data labeling. Prior to launch, Bark didn't just do a beta round but an alpha one, as well. In the first round, Bark worked with about 50 users and then in the second round progressed to a few hundred users from different areas to encompass a range of demographics. That "gave us great insight and confidence that algorithms were production ready," Bason said.

Bark's algorithms are based on "advanced machine learning and statistical analysis techniques to recognize potential problems," according to the company's website. Bason said that because human language is "so highly nuanced," relying on key words or even phrases is not sufficient to generate alerts for problems because the same phrases can convey different intents in different contexts. He offered the example of "I hate you." This sentence can be used just to express "standard envy or something more sinister," so you need the contextual clues, too. Bark's algorithms are designed to identify problematic messages because they have been trained with the millions and millions of conversations that are "fed" into the system. These messages teach the algorithm the contexts that enable it to accurately flag phrases that signify cyberbullying, sexting, and other potential problems.

Bark's alert levels are tailored to child's age. The company serves users from 5 to 17 years old, though the majority fall into the 8- to 14-year-old age range. Bark is not just for kids who already have their own phones, either, because many non-phone devices given to young kids are capable of messaging today, too. Bark offers "options to configure and customize a few factors like history of messages between participants."

Bark's approach seems to be working. Bason reports that more than 54% of children have experienced at least one issue that triggers an alert. Those alerts raise awareness of problems. Eighty percent of parents admitted to not having been aware of those issues before the Bark alert. It's all about that awareness, Bason said. Parents can't help their children if they are unaware of the problem.

Bason said parents come to Bark for a wide variety of reasons. In some cases, they come after their children have experienced some form of online bullying. Sometimes they come because they've seen these issues highlighted in the media. Parents may also learn about the dangers of social media from someone in their family or social circles. There are also parents who are buying their child their first device and want to provide some level of protection. Some of the parents may have attempted to use other products that they have found "too time consuming or too invasive."

(Image: Pixabay)

(Image: Pixabay)

Sometimes the child is the one who suggests the service. Bark doesn't give parents "unfettered access" to all a child's messages. It only shows parents the messages that were flagged as potentially problematic. The benefit for kids is that they don't have to sacrifice their privacy for the sake of safety. Bark also offers a way of "surfacing problems" for kids that they may feel uncomfortable bringing up themselves. The alert system means they don't have to report problems themselves, which takes pressure off them and can open the way for conversations with their parents.

Fostering these types of "healthy conversations" are a big part of the Bark solution. Aside from the alerts, the service says it provides advice on issues, including what practical steps to take, whether they are just to block an account or to go so far as alerting authorities. Bason said that parents consider the alerts "a great opportunity to talk about the risks that are out there." Reviewing the alert and the recommended actions constitute a "teachable moment" and an opportunity to get parents and children on the same page about online behavior.

Ariella Brown,

Ariella Brown is a social media consultant, editor, and freelance writer who frequently writes about the application of technology to business. She holds a PhD in English from the City University of New York. Her Twitter handle is @AriellaBrown.

Emojis Train AI to Recognize Sarcasm

It's sometimes difficult for humans to understand tone and sarcasm in text messages, so consider how difficult it is to train computers to understand it. But a new project is employing emojis to help AI understand when humans are employing irony.

Hiring Trend: Machine Learning Reduces Bias, Increases Applicant Pool

Unilever is using machine learning tools to increase the pool of entry level applicants and then test and evaluates candidates in the pool to cull those who are the best fits. Adding machine learning tools reduces bias, increases diversity, and makes the whole process more efficient, according to Unilever.


Re: Gimme that Old Fashioned Parenting
  • 3/19/2017 9:43:16 PM
NO RATINGS

..

Michelle writes "... I'm amazed at the number of parents who willingly hand over devices connected to the Internet without also providing guidance or some kind of monitoring."

Some of them also hand over a credit card, the keys to the house, and keys to the family truck ...

..

Re: Just my opinion on Social Media
  • 2/6/2017 8:58:05 AM
NO RATINGS

Yes, kids seem often to have the view that nothing bad will every happen to them, and that anything goes. It is up to parents and teachers to at least try to give some oversight and advice, even though there's still going to be some minority of those who either won't listen or don't want to listen to any guidance.

Re: Just my opinion on Social Media
  • 2/2/2017 1:36:35 PM
NO RATINGS

@PC indeed, credibility is earned and can be lost.

Re: Just my opinion on Social Media
  • 1/31/2017 8:39:26 PM
NO RATINGS

I think the saving grace for Twitter has been that many developers have used the API for practice in developing apps and now bots. I have heard other developers consider it relatively easy to use from a coding perspective.  Combine that with R programming and other advanced open source tools, and you can build a few great ideas in the process.

Re: Just my opinion on Social Media
  • 1/31/2017 8:37:40 PM
NO RATINGS

That "suckiness", for lack of better word, was probably ripe for machine learning. The volume of tweets, and then having to interpret intent, can be too much, even when you have a dedicated channel in Hootsuite or Tweetdeck to follow.  That had been my standard recommendation for clients - to look and see who you are following.  But the need now requires a more sophisticated approach. 

Re: Just my opinion on Social Media
  • 1/31/2017 6:22:36 PM
NO RATINGS

Keeping up with bullying and other inappropriate content is still a manual efffort for most including twitter. It's only going to get more complicated with live streaming. The key will be in developing more automated solutions.

Re: Just my opinion on Social Media
  • 1/31/2017 2:50:38 PM
NO RATINGS

@Ariella -

This is an old issue at Twitter. In 2015, then CEO Dick Costolo said "we suck at dealing with abuse and trolls"

With that history, I think I can be forgiven for answering the latest statements with "I'll believe it when I see it."

Re: Amazing Algorithm (s)
  • 1/31/2017 12:55:32 PM
NO RATINGS

Thanks Ariella for answering my question.  It is good to know they will be taking on other languages too.   Good Luck to them !

Re: Just my opinion on Social Media
  • 1/31/2017 12:37:32 PM
NO RATINGS

The challenge is for some bullying is subjective. Online bullying is usually manifested in person so something that may not appear that damaging online maybe coupled with in-person behavior that magnifies the online interaction. I understand the FLOTUS is taking on cyberbullying as her cause, so maybe some government funds will be targeted against these key issues.

Re: Just my opinion on Social Media
  • 1/31/2017 11:51:47 AM
NO RATINGS

@Broadway well, anything in writing may get into the wrong hands, but at least communication in a sealed envelope is better than a shout out across social media -- the equivalent of a post card that can be read by anyone who sees it along the way.

Page 1 / 6   >   >>
INFORMATION RESOURCES
ANALYTICS IN ACTION
CARTERTOONS
VIEW ALL +
QUICK POLL
VIEW ALL +