In the dynamic world of retail, two seasoned experts, Glynn Davis and Matthew Valentine, go head-to-head each month. This month, they spar over facial recognition!
IA et retail Les centres commerciaux Nouvelle mobilité
Facial recognition is another one of those areas that receives an incredibly bad press. The media goes into a frenzy whenever there is a hint that cameras are being used to identify individuals.
Let’s get this straight though. The fact is, most of the examples highlighted in the media are not actually using facial recognition but rather facial identification. There is a massive difference. One can identify an actual person whereas the other merely recognises there is a face and can suss out certain basic criteria such as approximate age and likely gender.
This was the case in Canada recently when the University of Waterloo had to remove vending machines from its campus because of some rumpus over them running facial recognition software. Nothing of the sort was actually taking place it transpired and the devices were merely undertaking local processing of digital image maps to recognise the presence of a person, their estimated age, and gender. And none of this data was then being stored.
What these types of devices invariably seek to deliver is a better level of service to customers but this is frequently forgotten in the frenzy of demonising any activity involving cameras and faces. The result all too often is that implementations are shutdown. This is a great shame because I’d argue that we are missing out on so many potential upsides from the technology.
Being recognised when you enter a fast-food restaurant and the kiosk serves you up your most frequently purchased items from previous visits would be a help for sure. However, even license plate recognition trials in restaurant drive-thrus have been met with an outcry. These initiatives clearly do not even involve a face!
One area that has been garnering great interest is protecting shopworkers from theft and violent attacks. But whenever cameras have been deployed to identify serial offenders then there has been much criticism of the businesses involved, which have included major players such as Co-op that has vehemently defended its actions and its employees.
Thankfully public opinion has been changing on this one and the calls of the British Retail Consortium (BRC) have been heard by the Government, which has agreed to invest £55.5 million in facial recognition technology over the next four years. This includes the roll-out of bespoke mobile units deployed in high streets across the UK to identify wanted people in crowded areas.
There is also a more fun side to facial technologies. Back in December Burger King in Brazil introduced a ‘Hangover Whopper’ initiative accessible via the company’s app that used facial recognition to determine the level of a person’s hangover (scored on the scale of one to three based on severity) and hand-out corresponding discount vouchers.
Are we to outlaw all this sort of activity alongside those implementations that provide massively effective solutions to some very serious problems? Let’s hope this is not the case. I’m not sufficiently naïve to fail to recognise that there are bad actors out there. This is sadly the case absolutely everywhere, but let’s not throw the proverbial baby out with the bath water.
There are myriad opportunities that can be derived from the use of facial recognition/identification that can benefit not only retailers and their customers but also wider society. Let’s deal with the technology using a measured approach rather than leaving it to a media-led witch-hunt. The latter all too often disappointingly prevails at the moment.
Sure, facial recognition technology could be a useful tool for retailers. But it could also be an Orwellian nightmare of invaded privacy, mistaken identity and unfairness. The decisions on which it is to be should be taken very carefully.
There is a natural appeal to the technology for retailers. Spotting shoplifters, recognising under-age customers seeking to buy alcohol or tobacco products, remembering people who have used forged currency… there are plenty of potential applications.
There are also plenty of hurdles to clear if the technology is to be used fairly. First and foremost is what many fear is an inbuilt fairness to the AI that powers facial recognition systems in the first place.
While the technology currently has a 99% success rate in recognising white, male faces it is far less accurate in recognising people of colour. Reports suggest it is especially poor at recognising black women.
Nobody should be subjected to restrictions on their use of shops because of mistaken identity, and if some groups are targeted more than others retailers will find themselves in a very difficult position – and one they might find impossible to justify. Adding racism or other discrimination is never a good plan for a company, and ‘the technology did it’ is no excuse.
But perhaps this technological challenge can be overcome. Perhaps then retailers can keep a digital record of faces, people they don’t trust or suspect of crimes.
Is that OK? What if they put your child on that database because their friend pinches a KitKat? What if a mishap with a self-checkout sees you put on that database? What if you have a doppelganger, or a twin?
These are important issues, not fringe ideas, and retailers should be cautious about how much of a fortress they make their premises.
Many retailers, and other groups such as local authorities, already discriminate against young people. For evidence, simply look into the number of ‘ultrasonic deterrents’ sold as ‘yob repellers’.
These devices emit a high pitched sound that is extremely unpleasant to young ears, though those of us who are older cannot hear them. These devices repel all young people, including those just trying to live their lives, or visit an area.
Would you trust these retailers to keep a database of faces, again possibly including your children? Would you know what happened with that data, what other security companies, landlords, retailers and organisations those images were shared with?
You wouldn’t, and you probably couldn’t.
