Yoti pitches biometric age estimation to regulate social media access for 7 to 11-year-olds
The biometrics-based age verification technology developed by Yoti is now likely accurate enough at determining the age of people between seven and 12 years old to support regulatory requirements for businesses to block people under 13 years old from social media apps or accessing content, CEO Robin Tombs writes in a LinkedIn post.
Tombs suggests that the combination of biometric liveness checks with AI age estimation will prevent almost anyone between seven and ten years old from joining social media platforms. While some 11 and 12-year-olds would get through, the number of 13-year-olds needing parental consent or to use the Yoti app would be limited.
People judged by the system to be 25 or more years old will be given the choice to have their age algorithmically confirmed, after which the image is deleted. People judged as 24 or younger can use a Yoti digital ID to share their over 18 status anonymously, allowing access to age-restricted goods at physical retailers, gambling, and online services like adult content or ecommerce websites.
The demographic breakdown of the age performance technology by Yoti’s July 2020 algorithm showed significant improvement over previous algorithms in performance for young adults, and Tombs noted at the time that the company expected its accuracy with 13 to 15 year-olds to improve as more people in the age group created Yoti digital IDs.
Indeed, the average MAE (Mean Average Error) rate for those 13 to 15 decreased from 1.71 to 1.60 in the latest results, and similar improvements were seen for other age groups. The overall average MAE fell from 2.46 to 2.28.
Yoti hopes to extend its shared results to two new age ranges, 7 to 9 and 10 to 12 by September of 2021, to provide regulators, businesses and parents with the transparent data they need to make informed decisions about protecting children. In order to do so, the company intends to build a new dataset of children’s biometric images.
Yoti launches two new campaigns for Safer Internet Day
The company also discusses the Age of Appropriate Design Code and how companies can provide age-appropriate services while preventing grooming in a company blog post for Safer Internet Day, which was February 9.
Yoti is participating in a sandbox being run by the UK Information Commissioner’s Office (ICO), along with partners like SaaS GoBubble (GoBubbleWrap) to continue the development of its privacy-preserving age estimation technology for children under 13.
The company is building a biometric dataset by asking parents to submit a photo of their child through a web portal, specifically for the purpose of training age estimation algorithms for improved performance. Yoti points out that the process is not facial recognition, as it does not match faces. It also does not reveal any personal information, but simply estimates age based on analysis by a neural network.
Yoti is also launching an education campaign and video competition to help parents and children understand AI and related concepts, including dataset consent and the differences between facial recognition and biometric analysis.
“The Marie Collins Foundation fully endorses the Share to protect campaign,” states Marie Collins Foundation CEO Tink Palmer, MBE. “We work with the victims of online abuse and know the damage caused to children. This initiative by Yoti needs to receive the full support of parents wherever they may live in the world.”