Confirming the age of children under the age of 18 is becoming an increasingly important requirement around the world in Australia, the UK, the EU and many US states, particularly for restricting younger children’s access to social media. Given children are less likely to have access to forms of ID or authoritative online records, concerns are often raised about whether this is practically deliverable. We would suggest a layered approach, combining both approximate age estimation and exact age verification methods, with a manual vouching process as a backstop to guarantee universal accessibility.
The first layer could be an estimation method, such as facial analysis. The state-of-the-art delivers results with a mean average error of +/- 1½ years, with under 1% of errors more than 2 years adrift with the real age of the user. So, on the basis that a 99% success rate is pragmatically agreed by policymakers and/or regulators, if the minimum age for opening a social media account is set at say 16, anyone whose age is estimated over 18 could reasonably safely be assumed to be old enough. Equally, anyone whose age is estimated to be under 14 has a less than 1% chance of actually being 16, so could safely be refused.
So, it is only those who are estimated to be within the margin of error for the estimation that we need to consider further. They could be asked to find an alternative method of proving their age. For many, they will have a bank account, a passport or even a driver’s license, so that is relatively straightforward. For those without any of these forms of evidence, ideally governments would facilitate access to authoritative records of children’s ages, such as those held by schools or social security benefits systems. These can be accessed using “one-way blind” checks, where an age assurance provider can only ask if data supplied to it by the user is exactly correct or not. The provider cannot see the data itself, and cannot check it multiple times in order to discern the information through trial and error. It can only get a thumbs up or a thumbs down. Depending on the nature of the data, an appropriate way to confirm the user supplying it is actually the person to whom it relates is also added. This could be, for example, another piece of information that only that person would usually have knowledge of.
There may still be some rare cases where a 16 or 17 year old, or even an adult, still has no suitable automated method open to them. In these cases, as a backstop, a manual vouching process can be made available where a trusted professional such as a teacher, JP or doctor, can provide a reference confirming the age of a user. To ensure that this is made available, an obligation needs to be placed on platforms to facilitate this more costly mechanism.
Having achieved the proof required for one social media platform, it is then possible through the interoperability ecosystem our partners at AgeAware® have developed, to re-use the same check when opening accounts with other platforms, so there is no need to repeat the process each and every time. There may be periodic authentication checks, to make sure the user who did the original check is still the same person trying to re-apply it, but that is a straightforward process as well, and will cause little inconvenience.
This layered approach will work at any age – 11, 13, 16, 18 or 65 – with the boundaries outside of which the second or third layers are needed simply being adjusted based on the expected accuracy of the estimation process.
Of course, some regulators in other sectors, such as online gambling, may still require a record of a confirmed date or birth, so the first estimation layer is not suitable and the process begins at the second layer. And users also have the option to skip the first step if they are concerned about sharing an image, when that is the estimation method on offer.
