City of Faces

When it comes to 21st Century surveillance, Connor Tomas O’Brien asks, what won’t we countenance?

Animation of facial recognition points lighting up

After my grandparents died, I helped to clear out their bookshelves at their home in Adelaide and found something that should have been discarded half a century earlier: a 1957 Sands & McDougall’s directory. Across more than a thousand pages, South Australians were organised into tidy columns, which you could navigate by address, name or occupation. 

I flipped to my grandparents’ street, and there they were. Or there, at least, was my grandfather – each home address had only one name attached – surrounded by neighbours frozen in time and place: salesmen, accountants, plumbers, fitters, graziers. How incredible and absurd, I thought, to be able to wander a city knowing who was behind each door. The White Pages doesn’t work quite this way – you can only search it by name – though it, too, now feels like a curio: I can find nobody I know in its latest edition, leaving it a ghostly, half-empty register of landlines long since disconnected. 

There is wonder in these proto-internets. The measure of a place was in the width of a spine. 

The real internet has made these objects redundant, but our changing norms have made them impossible in any case. Our conception of privacy swirls and cannot settle. Over the past few decades, it has rearranged itself profoundly. It’s worth considering that the very details we were, until recently, most comfortable sharing openly – our addresses and phone numbers, accessible to any stranger with knowledge of our surnames – are the pieces of personal information many of us now cling to most tightly, even as we broadcast so much else online. What’s perhaps even more remarkable is that we were ever comfortable sharing those details freely to begin with. 

There were no faces in the Sands directory, rendering the city it represented both radically open and entirely mysterious. Notions of public and private spheres were, in a sense, the reverse of our conception of those spheres today. While at home, you were publicly identifiable, but in public you could pass through the world anonymously. In an era when your name and address formed the core of your referenced identity, your face played no official part. Among strangers, it was entirely your own. 

Drivers licences were still issued without photographs as late as the 1980s – an era in which the federal government’s attempt to introduce photographic identification was met with overwhelming public opposition. If your physical appearance was officially noted, it was only in the crudest, most approximate terms: age, complexion, eye and hair colour. Until pretty recently, most Australians did not own passports. Officially, at least, your identity was separate from your appearance. 


John Brack’s Collins St., 5 pm, painted in 1955, reinforces the sense that, for much of the 20th Century in Australia, faces were not linked with public identity. An image that is constructed, in part, around the strangeness and singularity of faces – all drawn-out curves, contours, and sharp corners – seems simultaneously to reject the idea that individual faces signified anything in a public context. Nobody within the work is privy to the face of any other; only the uniform direction of movement – toward home, the site at which their identity is known – draws these strangers together.

Digital image of 'Collins St., 5 pm' by John Brack

John Brack, Collins St., 5 pm, c.1955

Collins St., 5 pm once served as a commentary on anonymity and stifling monoculture. If it were painted today it might read, at first glance, as a highly stylised (and culturally unrepresentative) interpretation of footage from a CBD surveillance camera. It’s unlikely that Brack himself would have foreseen this interpretation. Though George Orwell’s 1984 was published several years prior to Brack completing the painting, and East Germany was busily developing its surveillance state apparatus – it would be 36 years before any public space in Australia would be subject to camera monitoring. 

If I stand where Brack initially sketched the work, I can see at least six discreet cameras with me in their wide-angle view.’

The first Australian CCTV system was installed in Perth in 1991 and, as late as 2002, there were still major Australian cities with no public surveillance. The blanketing of recording devices across our cities is now almost complete. If I stand where Brack initially sketched the  work, I can see at least six discreet cameras with me in their wide-angle view. These cameras create their own reinterpretation of Collins St., 5 pm every evening. 

Our faces are now the focal points for public identity. If we had concerns about being watched, it feels decades too late to air them in any meaningful way. We’re already begun to take for granted the idea that cities are recorded spaces in which ownership of our faces is traded for right of access. It helps that CCTV systems have rapidly miniaturised, appearing to disappear exactly as they have proliferated. The same is true of cameras in general: metamorphosing from single-purpose objects to tiny bumps on the devices we carry with us at all times. (And just to put our modern surveillance states in perspective: it’s estimated the National Security Agency in the US has collected a billion times as much data as the Stasi collected in almost 50 years.)

If we’re accepting of cities as surveillance sites, that’s because we’re now much more willing participants in the de-privatisation of our faces. The internet, as a public space, has followed the same trajectory as our cities – from sites of semi-anonymity to panoptical places in which near-continuous access to our face is ceded as something close to an entry requirement. The two seem mutually reinforcing: our begrudging acceptance of social network tracking softens our ability to criticise increased real-world government surveillance, and vice versa. Our compliance with one form of monitoring ratchets up the other, or at least makes resistance seem pointless. 

Illustration of one of Brack's characters, subjected to facial recognition

By design, most surveillance feels benign, and that’s important when surveillance requires our ongoing participation. As in cities, when it comes to online spaces, it now feels as though it is our faces that are most of interest to those who are doing the surveying. We are, in general, more than willing to share – in part because the act of sharing our faces feels harmless, in the same way broadcasting our home address did not, until recently, strike many of us as inherently risky. 

There are good reasons to question why technology companies encourage our selfies. The most charitable interpretation is that feeds of faces are inherently addictive, and the more we supply, the more ad slots can be sold against them. This doesn't preclude our faces doing double-duty, though, as the raw input for systems learning to identify us. 

Facebook can now automatically identify us in photographs captured by those outside of our immediate circle – a feature which, despite being rolled out with little fanfare, is the product of a series of deliberately misrepresented technical breakthroughs. Over a decade, Facebook’s facial recognition system has evolved from offering shaky guesses about the identity of our close friends (picking from a line-up of a few hundred faces), to a system capable of, with high certainty, proactively seeking you out in pictures taken by strangers. If you were to upload a photograph of the once-anonymous throng heading home along Collins Street at 5pm any evening, there is now a reasonable chance Facebook could identify almost everybody in it. Not that the company needs to let you, or anyone it has identified in the image, know. It can be deliberately selective about which images of us it tells us it can see us in, to avoid the reputational damage it accrues with every privacy scare. 

The technological capability of facial recognition systems, at least in Australia, far outpaces the uses to which they can be put. This is what makes them seem benign, and encourages our continued participation in the addictive exercise of shooting, uploading, tagging. In truth, it seems likely that most companies with access to complex models of our faces – patchworks of different angles and expressions, sewn together from the array of photographs, videos and face-filter experiments we feed to Facebook, Google, FaceApp, Snapchat, Tiktok and others – still have no clear idea of how their systems will eventually be used. It is valuable enough for them, at present, to simply refine their models, waiting for a wildly profitable use-case to present itself.

Even as our surveillance system grows, our sense of shock is overridden by a feeling of resigned inevitability.’

Even as our surveillance system grows, our sense of shock is overridden by a feeling of resigned inevitability. One recent Facebook patent illustrates a system in which the cameras at self-checkout systems in retail outlets could be linked to its network, matching the faces of shoppers with those in its database. Is there any serious doubt this will happen, unless we radically change course? After all: if the cameras are already there, and we use Facebook’s services voluntarily, and we accept that access to our faces is the price of access to a space … every meaningful objection we have can be neatly swept away. 

It is notable that our most pointed fears around facial recognition tend to converge on China, a country in which such technology is increasingly being deployed for racial profiling and behavioural control. As others have noted, our critique of China may, in a sense, work to normalise the invasive surveillance apparatus that is being established right around us. A camera that can identify faces at the check-out might seem harmless compared to a system trained to identify and track faces belonging to members of a racial minority – but that does not mean that it is harmless. Our point of comparison matters. This is true on every scale. One technology company should not get a pass for being less nefarious than its competitor. The point of comparison should be the time when our faces weren’t routinely tracked. 


I still have the Sands directory. I have lugged it, for some reason, from house to house, and from state to state. The idea of a transparent, faceless city still seems strange – stranger, still, when you consider that these ideas of place and people have reversed so quickly, over just three generations.

We no longer broadcast our addresses. Pulling this knowledge out of the public domain has happened, in many cases, without us realising. We simply stopped receiving printed directories, or stopped responding to their requests for information. This new state of affairs works. It keeps us safe.

We now live in a city of faces. In one pocket, we may carry a wallet full of cards bearing our headshot. If our face matches those images we carry of ourselves, our identity is confirmed. In our other pocket, we carry a device for capturing faces, and sharing them to companies who will likely use them for financial gain.

Our faces are virtually impossible for us to change – at least, without great effort and expense. That’s what makes them great identifiers. Unlike our addresses or phone numbers, though, they may prove difficult for us to de-list.

Portrait of Connor Tomas O'Brien

Connor Tomas O’Brien was the Wheeler Centre's Digital Content Producer from May 2015 to June 2016. He was also inaugural director of the Digital Writers’ Festival and Digital Programs director at the Emerging Writers' Festival. He is also designer of Voiceworks magazine and co-founder of ebookstore platform tomely.com. He is working on a novel examining the nature of memory in our hyperconnected, hypermediated 'like economy'.

Discussion

All messages as part of this discussion and any opinions, advice, statements, or other information contained in any messages or transmitted by any third party are the responsibility of the author of that message and not the Wheeler Centre.