Privacy is under threat from the facial recognition revolution
Without protection, the rights of citizens and consumers will wither
We have grown used to cameras watching us. Once they recorded us; increasingly they recognise us, too. It is easy to think of facial recognition as just a technological convenience — a new way to unlock a smartphone, say. That is a mistake. In essence, facial recognition is biometric identification at a distance, with no need for consent.
Unlike fingerprints, retinal scans or blood samples, it is easily performed without the subject’s knowledge. It will affect how we travel, live, shop, and much else. It will force changes in the way privacy is defined and protected. If those who care about individual rights do not start thinking about the implications now, those changes will be forced upon us rather than chosen.
Facial recognition is already in use around the world. The Chinese equivalent of Amazon, Alibaba, allows people to “pay with a smile” using facial recognition in stores, for example. The potential for good is obvious. Think of the hours that could be saved if facial recognition were to become the default identification tool at airports.
These benefits will have to balanced against the loss of anonymity. In Russia, an app called FindFace identifies individuals in photos, linking them to profiles on a social network called VKontakte. A similar service, if linked to Facebook and other networks, could put names to billions of faces. In the city of Shenzhen jaywalkers are identified using CCTV, and their faces and addresses posted on a large screen to shame them into better behaviour.
The technology will not be limited to connecting a face with information already present on the internet. A facial recognition model developed at Stanford, when presented with paired photos of individuals who self-identify as gay or straight, could tell which was which with 81 per cent accuracy in men and 74 per cent in women. Humans given the same task were much less accurate. Yes, the sample was limited and the study needs to be replicated with a more refined methodology. The results cannot be dismissed, though. Nor can the frightening implications. Consider an algorithm identifying sexual minorities deployed in an intolerant, authoritarian state. The technology may misclassify many, but tyrants lose little sleep over false positives.
Facial recognition also may appear in a store near you, in the form of “dynamic pricing”. Supermarkets are replacing paper price tags with digital screens to avoid manually relabelling for price changes. Imagine if the prices changed based on the identity of the shopper, picked out by a camera at the store entrance. Does this sound too farfetched? Travel sites allegedly charge more when they detect the search patterns of a determined buyer. Amazon already changes prices quickly on the basis of demand. Economists call this price discrimination. With a set price, consumers who could have paid more capture surplus value. In a world of tailored pricing, this discount vanishes, increasing returns for sellers at the expense of buyers.
Dynamic pricing might be possible even without positive identification of the target. A University of Toronto study suggests that facial cues can provide useful hints about socio-economic status. A facial identification algorithm that detected class would be a hot commercial property.
Facial recognition technology will develop apace, whatever lawmakers and rights advocates do. When citizens have the right to anonymity is no longer a theoretical question. If the technology becomes ubiquitous before decisions are taken about legal limits, it may be the facts on the ground, rather than liberal principles, that determine the answers.