The days of showing your ID at a supermarket to buy age-restricted items such as alcoholic drinks may be coming to an end - thanks to facial age estimation technology.
A trial across a number of supermarket chains in the UK has shown promising results, with a facial scan working with artificial intelligence (AI) to determine whether shoppers needed to have their identification checked.
The trial, run by the UK’s Home Office, also allowed shoppers to try out another way of proving their age without any human interaction, with a digital ID app.
Yoti, the company behind the technology, said one of the key takeaways from the trial was that no underage customers purchased age-restricted items while using their age verification system.
The technology is better at estimating someone’s age than humans, the company added, reducing the risk of incorrect estimation, and reducing the need for those over the required age to carry ID with them.
Quick and accurate age verification
Shoppers who used facial age estimation technology during the trial looked at a camera on the self-checkout, and the algorithm looked at the face and conducted the estimation - a process taking around three seconds.
If it detected they looked younger than the set age threshold, a member of staff was alerted to check their ID.
According to the trial, around 70 per cent of shoppers said they would use the technology when buying restricted goods at a self-checkout, while the participating supermarkets said they would welcome legislative change to bring in digital age verification.
The UK government is trialling the technology, but it would need to change the laws on age verification for it to be rolled out outside of the trial.
Alongside Yoti’s technology in some supermarkets, the Home Office also trialled other forms of age-verification technology at a nightclub, a student bar, and an off licence.
How does it work?
Yoti uses machine learning to build a neural network that is trained on images of the human face and accompanying age data.
In its latest White Paper, the company states its technology cannot identify an individual, and immediately deletes all images of users.
It can estimate the age of people aged 13-17 as under the age of 23 with 99.65 per cent accuracy. For 6-11 year olds, it estimates their age as under 13 at 98.91 per cent accuracy.
Yoti therefore believes this technology has a number of different use cases besides age verification at physical checkouts, such as verification for digital products.
The company claims it has minimised gender and skin tone bias within its algorithm, although its data shows there is slightly more mean estimation error for females over males, and for darker skin tones over lighter ones.