It’s awful when children are bullied at school, but when it happens online, it’s even harder to escape. How can parents protect their children on social media without violating their privacy?
Surfacing the warning signs
Using cloud-based AI from IBM Watson, Guardio built a solution that monitors messages on WhatsApp and Instagram, and alerts parents about exchanges that could indicate bullying.
<0.01% rateof false-positives ensures only high-risk messages are sent to parents
Alertsparents to help victims get the support they need
Rapidtime-to-market enabled by pre-built cloud-based IBM solutions
Business challenge story
Bullying: an old problem with a new face
Bullying is an age-old problem, but the advent of social media has dramatically changed its shape and severity. Around 25 percent of ten-year-olds have smartphones, and as children interact online, bullying has moved from the schoolyard to the internet. This move to the online space can have significant consequences, as victims struggle to escape abuse. What’s more, many perpetrators are more aggressive online than face-to-face, because they do not see the victim’s reaction and do not feel empathy.
Recognizing the ever-growing threat that cyberbullying poses to young people, Pablo Elosua co-founded Guardio and stepped into action. He understood that many parents struggle to protect their children against online harassment, particularly because the warning signs of cyberbullying can be hard to detect.
Pablo Elosua, now CEO of Guardio, remarks, “There are some tools available that help parents control what their children encounter online. However, many existing methods rely on blocking specific sites, or allowing parents to read their children’s messages – invading their privacy. Additionally, with the volume of messages and sites, parents become overwhelmed. We wanted to find an effective way to protect children while respecting their privacy, and we needed a platform that we could scale to support thousands of users as we grow.”
Surfacing the warning signs
The Guardio team realized that text analysis could provide the answer. While free software appeared to offer a solution, it needed significant investment in development. Instead, Guardio considered taking a cloud approach, which would enable a start-small scale-fast launch, and looked for ways to reduce the time to market.
Guardio joined the IBM Global Entrepreneur program and began using IBM Cloud™ and IBM Watson™ technologies. The advanced IBM Watson services offer sophisticated text analysis, greatly reducing Guardio’s training and investment requirements. Based on the existing core capabilities of IBM Watson, Guardio moved from concept to production extraordinarily rapidly.
So, how does the new solution work?
Guardio’s service relies on IBM Watson Natural Language Classifier, which sifts through thousands of everyday exchanges to pinpoint likely cases of bullying. For rich textual analysis, Guardio harnesses IBM Watson Natural Language Understanding to capture the sentiment and tone of each message – for example, determining whether an insult is being used in a harmless or aggressive manner. To help the IBM Watson solutions to understand the specific vocabulary and colloquialisms used by children, volunteers submitted existing conversations to build up a library of expressions. In action, Guardio analyzes conversations in full rather than individual messages in isolation – helping to ascertain crucial factors such as the density of insults. IBM Cloud delivers the Watson APIs.
The service currently supports WhatsApp and Instagram, which between them cover 80 percent of social media activity for people under 14. Parents connect their children’s social media accounts to the Guardio service using a QR code, then Guardio automatically analyzes children’s online activity, giving each conversation a risk score based on its likelihood to exhibit bullying. Conversations that exceed a certain risk threshold are sent to the parent via an automated email notification, and all other messages are not even stored in the data base.
Pablo Elosua explains, “The IBM Watson solutions are very easy to implement and use, and its language understanding capabilities have enabled us to move to market very rapidly. With IBM Watson handling text analysis, we have been able to focus on connecting to social media and developing the protection that matters to parents. From concept to execution, Guardio developed and launched its first service in six months, and without IBM Watson solutions it would have taken much longer.”
Offering effective protection
Guardio’s service analyzes around 2,000 messages per day, and identifies likely cases of bullying with high accuracy. With the solution in action, Guardio focuses on promoting the service and acquiring new users, and with the solutions based on IBM Cloud, Guardio can easily scale up to support volume. As the system grows in popularity and starts analyzing more messages, it will further refine its algorithms via machine-learning, helping to detect cyberbullying with even greater accuracy.
“Understandably, many parents are keen to see that the system is effective before they start using it, so they send sample messages, and observe whether the solution picks up on aggressive communication,” continues Pablo Elosua. “The feedback from parents has been excellent, and we are delighted that IBM Watson’s capabilities give parents peace of mind that their children are safe online.”
One of Guardio’s major concerns was false-positives – for example, wrongly identifying a child’s normal conversation as malicious and notifying their parents. IBM Watson services distinguish effectively between harmless and aggressive messages, and the solution’s rate of false-positives is below 0.01 percent.
Pablo Elosua concludes, “The scalability offered by taking this approach with IBM Cloud means that we can grow as the opportunity arises. Eventually, we want to use the same IBM Watson APIs to help detect and address other issues stemming from children’s exposure to online material, such as eating disorders, grooming and radicalization. Our platform has so much potential, and with IBM Cloud and IBM Watson we are excited to see what we can achieve in future.”
Guardio is a not-for-profit start-up that offers a free, AI-based service to help parents protect their children against bullying on social media. Founded in 2016, the company is based in Madrid, Spain.