The Classics in your pocket
Thanks to Sir Francis Galton, and now AI, ancient coins, it turns out, still have modern face value
Does your phone, like mine, unlock at merely a glance? Do you remember the last time you went through Border Control and had to look into that camera? AI was scanning your face, turning your features into data using an algorithm trained disproportionately using white male faces and comparing them to a composite or authoritative image of you it has on file. So you just look at your phone and hey presto, you are in, texting away. Amazing. But those amazing facial recognition algorithms had to be taught how to handle faces, and to learn that they needed data-lots of faces. Now where do you think early developers found publicly accessible data sets of millions of faces? Can you guess?
As Senior Principal Researcher at Microsoft Research Kate Crawford has argued, the answer is drawn from US Government records: application, visa, and border crossing photographs, but most notably those scraped from law enforcement, mug shots. They are made available for facial recognition algorithm testers on the internet as part of the National Institute of Standards and Technology’s (NIST) Special Database 32-Multiple Encounter Dataset.
Now mug shots have a history as old as photography, as Crawford and others have shown. They were born way back in the nineteenth century, from two sets of fathers: one was the criminologists like Alphonse Bertillon, who wanted a way of spotting individual repeat offenders, and who by devising a classification system which combined biometrics (bodily measurements) and side and frontal facial photography in late 1870s Paris, pioneered the identity card (Fig. 1); but the other was the European so-called ‘physiognomists’ who set out to theorize the ‘criminal type’.
Building upon Aristotelian contentions about the possibility of sympathetic correlations between physiognomy and character and their development in antiquity and beyond into a science of physiognomy, the nineteenth-century physiognomists were obsessed with relating what people looked like with what they could or would do, their innate capacities and propensities, and with trying to figure out what made people typical criminals. (The Italian Cesare Lombroso’s L’uomo delinquente, The Criminal Man, published in 1876, was a notable example that suggested the signature of criminality was written in the structure of the face, and one paralleled in Britain by the work of Sir Francis Galton).
Sir Francis Galton, a racist British polymath and correspondent of Bertillon, said that it all boiled down to breeding — to what he called eugenics 'the art of good breeding' — and the particular character allegedly expressed in the proportions of certain facial features. But he questioned the naivety and subjectivity of judgements of individual physiognomy. In 1878 he famously came up instead with the idea of a 'ghost image' made from a composite of photos of many criminals from which he tried to essentialize the 'criminal type'.
And to corroborate his new method and the significance of its findings, intuitively he turned — like many other nineteenth-century racists — to the artefacts of antiquity. Compiling composite photography of dozens of ancient coins featuring Alexander the Great, Nero, Cleopatra VII (whose features, were in sum, he said — despite her alleged beauty — ‘simply hideous’ ‘to ordinary English tastes’), and others (including some ideal Roman female likenesses ‘singularly beautiful’ in composite)— coins that were in their own time statements of power — he compared his ‘criminal type’ with the racial and characterological profiles that famous ancient faces allegedly exemplified (Figs 2-4). You see, here, for Galton (as his colleague Flinders Petrie had also argued of Greek sculptural faces in his cognate photographic explorations) was to be found a visual catalogue of superlative character, a facial record of the inner quality of ancient rulers, and 'the ablest race', as he called the Greeks (although neither Galton nor Petrie realized that all those white Classical busts were once brilliantly coloured): perfect Grecian noses (as the physiognomists saw them), the facial angles of the most fully realized rational men (steep not sloped) not of the ‘lower races’ or ‘criminal types’, the visage of hereditary genius, and so on. And so bolstered by the evidence of antiquity, one iteration of the mug shot as data was born: the translation of profile and frontal photos of individuals who Galton suspected were biologically fated to commit crimes into generalized data so that others with characters just like them could be spotted.
Well, as photography expanded all over the world as just one of many European colonial technologies of classification and control so did the mug shot. And alongside the identity card’s individual portrait, it became the archetypical criminal image - the photo that could be released to the Press even before anyone went to trial and was proven guilty; so all the way through the history of modern journalism, from the nineteenth to twentieth centuries to today, you don't see family photos of a suspect in the Press, you see his or her police mug shot. And despite the twentieth-century debunking of the physiognomists’ determinist claims, thanks to that and to Galton and his peers, what you see is the suspect as criminal. You see the very picture of guilt.
Until now, when, in search of millions of faces from which to scrape data for their algorithms, the AI developers turned to the NIST (now part of the US Department of Commerce) and to the mug shot, testing their twenty first-century iterations of Galton’s generalizing method of mapping facial standards in order to develop not a typology of the criminal, but a new technology of discrimination: the amazing technology of facial recognition. And their eyes biased just like Galton’s, towards people who look just like them (principally white males), disproportionately represented in their data sets, perhaps it's not too surprising that their new technology of discrimination, too, struggles to recognize women and people of colour. So when you carry around your phone with its algorithms honed to detect the geometry (or pixels) of your face, unnerving to think that perhaps bouncing along in your pocket sits a little bit of recycled racist history, the ‘ghost image’ of one reinvention of the Classical past, now hidden in plain sight, that still dares brazenly to look back at you when you reach down to check your phone.
Figures 1-4, left to right
1: Fiche anthropométrique d’Alphonse Bertillon, 11 juillet 1894. (Source: Wikicommons images)
2-3: Plates from: Galton, F. (1879) ‘Generic images’, Proceedings of the Royal Institution 9: 161-70
4: Galton, F. (1883) Inquiries into Human Faculty and its Development. Macmillan: Illustration plates, p.7 (Source: Wikicommon images)
Ashley Clements
Ashley Clements is Associate Professor in Greek literature and Philosophy & Head of Department of Classics, Trinity College Dublin.
His most recent book is Humans, among Other Classical Animals (Postclassical Interventions). Oxford University Press (2021)