What is age estimation technology?
In brief, age estimation technology is the ability to estimate a person’s age using technology. Where traditional facial recognition technology is used to identify an individual, age estimation is only interested in confirming their age.
Although the two technologies seem similar, there are important distinctions between them.
As Instagram introduces age estimation technology to verify the age of teenagers, between 13 and 17 years old, who use their platform, it begs the question – what is this technology? And how does it work?
Facial recognition technology works by separating a person’s face from their background imagery. The separated facial image is then checked for details such as facial posing, grayscale, and illumination. Features like eyes, nose and mouth are then measured. These extractions represent a digital face, and that face is compared to a database. When an organisation needs to check a person’s age, they would traditionally need personal information or credit card details for that person – and that information can sometimes be checked against third parties, such as credit reference agencies. More importantly, the technology is designed to identify an individual. However, the technology is surrounded by controversy because it isn’t always accurate. It can also gather data online without permission - The Information Commissioner’s Office (ICO) fined Clearview AI in 2021 for failure to comply with UK data laws when it found the company used peoples’ online images without permission.
Age estimation technology uses the same principals as facial recognition technology in terms of pinpointing facial details, but it does not ask for personal or identifiable information, so the user remains anonymous.
Instagram announced plans to use a programme called Yoti to perform its age verification analysis. Yoti’s system has been built in-line with GDPR’s ‘data protection by design and by default’, previously known as ‘privacy by design’ – meaning that an individual cannot be identified, and only the minimum data is needed to carry out the process because it only needs the person’s facial image. Once a person’s age is verified, their image will be immediately deleted from its database.
Yoti’s estimation technology is performed by a ‘neutral network’ which Yoti claim can identify a person’s age to within 1.36 years of 6- to 12-year-olds, and within 1.52 years of 13- to 19-year-olds. Its technology uses a form of artificial intelligence (AI) which uses an age estimation algorithm to check millions of images to estimate how old an individual is.
Who is Yoti?
Yoti is a digital identity app that uses age estimation technology and provides privacy-friendly technology to organisations who need to check their customer’s age.
Age estimation and facial recognition systems are categorised as ‘biometrics’ because they involve the measurement of a human’s physiological characteristics.
Biometrics work by:
- Detection – is there a face in the picture?
- Characterisation – What assumptions can be made about this face?
- Unique persistent identifier – What is this person doing, in a limited context, not tied to other PII? (personal identifiable information)
- Verification 1:1 – Is this person who they are claiming to be?
- Identification 1:Many – Can software determine who this unknown person is?
Once a person’s age has been identified, social media platforms can then tailor the users online experience to age-appropriate content and implement safety controls such as turning off excessive notifications, setting geolocation (but the child has the option to disable this), minimising data collection, using child-friendly language and making sure that they are treated appropriately for their age.
But this is only part of the issue, and it poses the question, why do children feel they need to pretend to be older online in the first place? As more people join the Metaverse, there are concerns over privacy and data collection – such as the extraction of personal information, the lack of authority in the space and its unknown reach and possibilities. Also, the potential for criminals, hiding behind an avatar, to prey on vulnerable people is concerning. The space has been described, by VR users, as an “unsafe wild west”, and it is estimated that children will spend 10 years there so it is important that safety measures are in place.
So could age verification technology, by tailoring experiences to age appropriate content, help to police the space so that children are safeguarded?