Is age-estimation technology the future of digital safety? INHOPE interviewed Julie Dawson, Chief Policy and Regulatory Officer at Yoti about their technology. Why is it now more relevant than ever before, and how will it affect the future of our digital landscape?


Modelling the UK's age-appropriate design code, which outlines legal guidelines online platforms have to follow to ensure safe experiences for young users, the state of California passed a bill to protect kids online. The code being implemented on the West Coast of the United States (where most of the big tech headquarters are located) signals a shift within the digital safety landscape. As a result, global policymakers are now increasingly recognising the importance of facilitating a safer digital experience for young users through legislative action.

Up until the recent legislative changes, creating safe digital experiences for kids was a topic almost exclusively discussed within the field of Child Protection. "Now, with changing legislation the discourse is slowly shifting and gaining societal relevance," says Julie Dawson. Previously, most online platforms were only required to specify a minimum user age in their terms and conditions. In some jurisdictions, they are now becoming legally obligated to also enforce these terms and can face high penalties when that is not ensured.

But how can companies reliably verify user ages without collecting their data and interfering with their privacy? A potential solution to this is using age estimation technology. We spoke with Yoti to understand more about how it works, and how their technology can help to keep young users safe online.

How does it work?

Yoti's technology estimates the probable age of a person by analysing an image of the user. Different from face recognition, the system is not trained to recognise a person, but to calculate the probable match between the age of the person in the image and the patterns learned from the training data. After the age estimation is performed, the captured facial image is deleted from the backend servers and the estimated age is relayed to the organisation. No images are stored, shared or used for any other purpose than estimating the age.

What differentiates Yoti's age estimation from other age verification processes, is that the product was created with the main focus on user privacy, security, and inclusion explained Julie. "Our motivation is to both be data minimised and enable the platform to meet age-appropriate design requirements." The user does not have to register or provide personal information to use the technology, they simply have to present their face in front of the camera. The process is based on Artificial Intelligence (AI) which has been previously trained on a large number of annotated images of faces. To the technology, the image is simply a pattern of pixels and it has been trained to spot patterns in numbers. It is accurate to 1.52 years (Mean Absolute Error) for 13-19-year-olds and 1.56 years for 6-12-year-olds.

Why do we need age estimation?

Currently, not all online platforms have a reliable method to confirm their users' age besides physical Identification documents, such as a credit card or driver's licence. "Over a billion people in the world do not have a government-issued birth certificate, passport, or driving licence," pointed out Julie, "so what are ways for people who either don't own or don't have access to their documents to use those digital platforms?" Age estimation technology allows tech companies to create age-appropriate digital experiences for everyone between the ages of six and seventy.

In many countries, children may not have a legal form of ID. In these circumstances, age estimation enables social media and gaming platforms to reliably assess their young users' ages and create age-appropriate experiences for kids. "Children of different ages will need different protection and support on a given platform," explained Julie. The age of a young user can give companies the insights to make an informed decision on which steps can be taken to ensure a safer and appropriate experience for the child. Yoti proposes several examples of potential age-dependent measures:

• Turning off age-inappropriate advertising
• Minimise data-collection
• Disabling chat or camera options
• Enable age appropriate Content Moderation
• Disabling geolocation
• Providing support and digital education resources in child-friendly language
  
  

How reliable is this technology?

Ensuring a safe and age-appropriate experience for all young users is what all of us have been striving for, but how dependable is age estimation and are there ways it can be manipulated? We asked Julie Dawson about potential threats to the tool. "There are lots of different attack factors, such as young people trying to look older, or older people trying to look younger," she explained. To avoid risk, Yotis technology applies a buffer of three to five years, when a minimum age requirement is a legal obligation, such as for purchasing alcohol or visiting adult sites

In certain cases, however, such as on most social media platforms minimum age is not a legal requirement. "We have to look at what's proportionate" noted Julie. For sites that do not yet have a legal obligation to verify age, the implementation of age estimation is an easy but big step in improving safety for young users.

The Future of Age Estimation Technology

"We know that the accuracy of these technologies is increasing", said Yoti's Chief Policy and Regulatory Officer; "Organisations are warming up to the age estimation approach, as it is scalable and affordable and customers are increasingly choosing age estimation over other age verification processes." Age estimation seems to be one of the least disruptive methods of implementing age verification on a platform that previously did not require it. This characteristic could make age estimation one of the defining features of an evolving digital safety landscape.


Regardless of which technology we will be using in the future, we must understand the importance of providing age-appropriate experiences. To keep children safe in real life we have protections and legal age requirements in place that, while not faultless are more or less easily enforced. Implementing these requirements online is a bigger challenge. Age estimation technology could be the answer to make this process more approachable in the future.

You can learn more about Yoti and its age estimation tool by clicking here.