Predator-catching app is using algorithms to detect children’s emotions
so it can alert parents when online threats surface
Sophie is an iPhone app developed by researchers in South Australia.
Sophie uses artificial intelligence (AI) to alert parents when their children demonstrate emotional behavior such as anxiety or fear. The app is centered around a third-party keyboard, which studies typing patterns to identify when a child is under duress while on the device.
It first builds a psychological profile of a child to use as a base to identify abnormal reactions.
Co-founder Ben Flink said he came up with Sophie after hearing about a friend’s child who had been contacted by an online predator through a computer game.
He said there was a glaring need for better safety features on mobile devices because more children were logging on to the internet at a younger age with minimal protection or awareness of the risks.
“There are 17 million kids aged 12-17 years old on the internet on their mobile devices in the US alone and 10 per cent already admit to physically meeting up with strangers,” Flink said, “Predators like to elicit specific types of emotional and psychological responses from a child to build up a relationship and disenfranchise them from their existing network.
“Sophie uses machine learning to pick up on emotional cues and through its algorithms can alert the parent with an indicator that perhaps the child is talking to a predator.”
According to the National Center for Missing & Exploited Children, there are almost 850,000 registered sex offenders in the United States.
Sophie is intended to be a safety net for children who are first-time smartphone users.
Similar to the way Fitbit watches track physical health and movement of the user, Sophie tracks language input on a Smartphone and monitors the mental wellbeing of a child.
The app does this by studying the intensity, timing, tone, and language of the user when they type and uses the information to determine a baseline of normal behavioral activity.
Using a number of languages and behavioral algorithms, the app searches for any significant variations in those categories.
If there are any spikes of activity outside the child’s normal emotional range, an email is sent to the parents notifying them of the abnormal behavior. It provides them with relevant advice on how to address the situation.
The app also can be used to report cyber crimes including child exploitation and cyber bullying to authorities.
Flink said although other anti-predator apps focused on locking, blocking and filtering a child’s phone, children and predators could still work out ways to get around these approaches.
He said Sophie detected and analyzed all language input through any app on the phone, making it a more efficient and less restrictive protection technique.
It can also help to protect children from online bullying and mental health issues.
“It’s not just looking at key words or phrases but a whole input of behaviour,” Flink said, “As someone with a young family, knowing that the current tools aren’t working means something needs to be done, “We want to use Australia as our test market at the moment before heading into North America because this is a global issue.”