Physiognomy’s New Clothes

Figure 1. A couple viewing the head of Italian criminologist Cesare Lombroso preserved in a jar of formalin at an exhibition in Bologna, 1978. (Photo by Romano Cagnoni/Hulton Archive/Getty Images)

Introduction

Machine learning for understanding images

Figure 2. Image dating with deep learning. ChronoNet guesses 1951 (left) and 1971 (right).

Learning a “criminal type”

“[…] 330 are published as wanted suspects by the ministry of public security of China and by the departments of public security for the provinces of Guangdong, Jiangsu, Liaoning, etc.; the others are provided by a city police department in China under a confidentiality agreement. […] Out of the 730 criminals 235 committed violent crimes including murder, rape, assault, kidnap and robbery; the remaining 536 are convicted of non-violent crimes, such as theft, fraud, abuse of trust (corruption), forgery and racketeering.”

“non-criminals that are acquired from Internet using the web spider tool; they are from a wide gamut of professions and social status, including waiters, construction workers, taxi and truck drivers, real estate agents, doctors, lawyers and professors; roughly half of the individuals […] have university degrees.”

“[…] the angle θ from nose tip to two mouth corners is on average 19.6% smaller for criminals than for non-criminals and has a larger variance. Also, the upper lip curvature ρ is on average 23.4% larger for criminals than for noncriminals. On the other hand, the distance d between two eye inner corners for criminals is slightly narrower (5.6%) than for non-criminals.” [7]

Figure 3. Wu and Zhang’s “criminal” images (top) and “non-criminal” images (bottom). In the top images, the people are frowning. In the bottom, they are not. These types of superficial differences can be picked up by a deep learning system.

“[…] participants, given a set of headshots of criminals and non-criminals, were able to reliably distinguish between these two groups, after controlling for the gender, race, age, attractiveness, and emotional displays, as well as any potential clues of picture origin.”

Figure 4. Stereotypically “nice” (left) and “mean” (right) faces, according to both children and adults.

“We are the first to study automated face-induced inference on criminality free of any biases of subjective judgments of human observers.”

Scientific racism

Figure 5. Like man, like swine: From Giambattista della Porta’s De humana physiognomonia (Naples, 1586).
Figure 6. Francis Galton’s attempt to reconstruct an “average criminal face”.

“Intelligence, activity, ambition, progression, high anatomical development, characterize some races; stupidity, indolence, immobility, savagism, low anatomical development distinguish others. Lofty civilization, in all cases, has been achieved solely by the “Caucasian” group.”

Figure 6. The idea that there are inferior types of humans has historically been linked to the scientifically invalid idea that some humans are more like animals than others. From Nott and Gliddon, Types of Mankind, 1854.

“[…] man bears in his bodily structure clear traces of his descent from some lower form; […] [n]or is the difference slight in moral disposition between a barbarian, such as the man described by the old navigator Byron, who dashed his child on the rocks for dropping a basket of sea-urchins, and a Howard or Clarkson; and in intellect, between a savage who does not use any abstract terms, and a Newton or Shakspeare. Differences of this kind between the highest men of the highest races and the lowest savages, are connected by the finest gradations.”

“Just as it is often difficult to tell a toadstool from an edible mushroom, so too it is often hard to recognize the Jew as a swindler and criminal […] How to tell a Jew: the Jewish nose is bent. It looks like the number six […]”.

Figure 7. From Vaught’s Practical Character Reader, 1902, p. 80.
Figure 8. Nazi “race scientists” doing institutionalized physiognomy, 1933.

“There’s evidence (re)emerging […] that a person’s looks do say something about his politics, smarts, personality, and even his propensity to crime. Stereotypes don’t materialize out of thin air, and the historical wisdom that one can divine the measure of a man (or a woman) by the cut of his face has empirical support. […] You CAN judge a book by its cover: ugly people are more crime-prone. […] Physiognomy is real. It needs to come back as a legitimate field of scientific inquiry […]”.

“Faception is first-to-technology and first-to-market with proprietary computer vision and machine learning technology for profiling people and revealing their personality based only on their facial image.”

Unexamined assumptions

Figure 9. From Dorothea Lange’s “Migrant Mother” series. The original caption reads: “Destitute peapickers in California; a 32 year old mother of seven children. February 1936.”
Figure 10. Stimuli in R. Jenkins et al’s 2011 paper Variability in photos of the same face.

“The spirit of Plato dies hard. We have been unable to escape the philosophical tradition that what we can see and measure in the world is merely the superficial and imperfect representation of an underlying reality. […] The technique of correlation has been particularly subject to such misuse because it seems to provide a path for inferences about causality (and indeed it does, sometimes — but only sometimes).”

Criminality

“not so much “brutalized” (in the modern sense: deformed by ill-treatment) as he was “a brute,” whose criminal nature was written on his very skin.”

“[…] the truly durable legacy of the convict system was not “criminality” but the revulsion from it: the will to be as decent as possible, to sublimate and wipe out the convict stain, even at the cost […] of historical amnesia.” [19]

“What do you hope to conclude from the similarity of faces, especially the fixed features, if the same man who has been hanged could, given all of his dispositions, have received laurels rather than the noose in different circumstances? Opportunity does not make thieves alone; it also makes great men.”

“In pairs of either natural or composite faces the face higher in testosterone was chosen as more masculine 53% and 57% of the time respectively. The authors argue that only men with very high or very low levels of testosterone may be visually distinguishable in terms of their masculinity. […] other studies find no links between testosterone and masculinity. A study using almost identical methods […] but with a much larger set of men, found no association between perceived facial masculinity and testosterone levels […] Similarly, Neave, Laing, Fink, and Manning (2003) reported links of perceived facial masculinity with second-to-fourth digit ratio (2D:4D), but not with measured baseline testosterone levels; and Ferdenzi, Lemaître, Leongómez, and Roberts (2011) found no association between perceived facial masculinity and 2D:4D ratio.”

“It sucks to be poor, and it sucks to feel that you somehow deserve to be poor. You start believing that you’re poor because you’re stupid and ugly. And then you start believing that you’re stupid and ugly because you’re Indian. And because you’re Indian you start believing you’re destined to be poor. It’s an ugly circle and there’s nothing you can do about it.”

“Unlike a human examiner/judge, a computer vision algorithm or classifier has absolutely no subjective baggages, having no emotions, no biases whatsoever due to past experience, race, religion, political doctrine, gender, age, etc., no mental fatigue, no preconditioning of a bad sleep or meal. The automated inference on criminality eliminates the variable of meta-accuracy (the competence of the human judge/examiner) all together.”

Conclusion

Thanks

Notes

--

--

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store