Precis of face reading paper

Computer vision researchers are beginning to tackle the highly salient field of phrenology. Every human culture has believed that inferring character from physical features is possible. Whether humans are merely superstitious or employing useful mental heuristics, they tend to agree with each other about which faces look dominant, attractive, trustworthy, or extroverted. This paper recreates this ability with objective computational pattern-matching, and the results generally agree with human intuition.

The particular concern of this paper is to build a computer program that predicts whether a person is a criminal or an ordinary citizen, using only an image of their face. Though it ought to be fairly easy to distinguish “normal”-looking faces from abnormal ones, this subject has received little recent attention due to its historical association with the horrors perpetrated by adherents to social Darwinism. The objectivity of electronics eliminates problems like ideological bias and human incompetence, and the field of computer vision has finally reached a level where computers can match or exceed human abilities in facial recognition.

This paper uses standard ID photographs rather than computer-generated face models, which would not be representative of actual criminal and non-criminal populations. The dataset consisted of 1,856 Chinese men, 1,126 of whom were non-criminals. Many aspects of the photos were normalized to avoid confounding variables, such as lighting intensity.

Four machine learning methods were employed: Nearest Neighbor, Logistic Regression, Support Vector Machine, and Convolutional Neural Network. The first three of these rely on a vector of facial landmark points, a PCA vector thereof, an LBP vector thereof, and finally a concatenation of these vectors. Each of the four methods was then tested several times against random samples from the population and the results were averaged to yield its predictive power.

The Convolutional Neural Network had extraodinary predictive power, averaging 89.51% accuracy, and the others performed with 78% accuracy or better.




(This takes a bit longer than I expected. To be continued…)

About Aeoli Pera

Maybe do this later?
This entry was posted in Uncategorized. Bookmark the permalink.

12 Responses to Precis of face reading paper

  1. Edenist Whackjob says:

    Machine Learning > Javascript, in terms of job prospects. Fewer jobs, yes, but more favorable supply/demand ratio.

  2. Boneflour says:

    Physiognomy Is Real™

  3. Lowballer says:

    I don’t think this one would work very well for Edenism. It suggests people with Neanderthal features are more likely to be criminals, while Sapiens are more law abiding (although that doesn’t mean much in a lawless wasteland like China).

    • Aeoli Pera says:

      This is true. But what I think Laz meant is, this would be the page 1 proof of concept under the heading “Physiognomy Is Real™”.

    • bicebicebice says:

      “It suggests people with Neanderthal features are more likely to be criminals, But that is true tho, in a sense, we follow moral laws and saps follow Melon ruling, in a nutshell.

  4. Anthony S. Culver IV says:


    I’m not so sure we can chalk these features up to Neanderthals. The report seemingly suggests that wide interobital distance is a feature of non-criminals. Neanderthals seem to have wide interobital distances, albeit less so than Erectus and earlier hominids. There is conflicting data about modern human interorbital distance. Howells (1972) gives a range of 18-24mm for modern humans, Neanderthals have a range of 23-32mm. Negroids and Australoids are apparently at the upper limits of of modern human variation in this character and fall within Neanderthal values. Yet this a pitfall for this study, and calls in to question the claim that is adjusts for racial differences. If narrower interorbital distances are assosciated of criminal status, why are Negroids and Australoids disproportionately imprisoned relative to Caucasoids?

    And then there’s this mouth-nose angle. Is there any evidence Neanderthals had narrow mouths? Their palates are several orders broader thsn modern humans, as are their mandibles and maxillae. They’re also sometimes quite prognathic which can stretch lips around considerably. I see no reason to conclude that this is a Neanderthal trait. A narrow mouth is a Caucssoid trait.

    Finally: the lip curvature. We have no evidence Neanderthals had excessively curved lips. This is a soft tissue feature that isn’t preserved on skulls, and it seems there is little in the way to infer about whether their lips were curved. Lip curvature is also supposedly a Caucasoid trait that Negroids and many other races with disproportionate criminal convictions are lacking.

    All in all, this study seems highly dubious and I doubt it would have made it past peer review in a Western country. I admit that, in looking at the following photo, one is inclined to assume that the criminal faces are more Neanderthal-like, and that the non-criminal faces are less Neanderthal.

    One problem that hasn’t been resolved is: how did the authors normalize the photos for size? There is no way of doing this without knowing exctly what kind of lens the camera was using and the distance of the lens from the subject. It is highly unlikely the authors did this for the +1,000 photos. Since we are talking about measurements that vary as little as 5mm, this is critically important. The quality of the pictures is so low that it looks like merely skipping a pixel or two could throw off an accurate measurement.

    China is notorious for its lack of human rights and general scuminess. You are right to point out that many people in China would be criminals if China had anything resembling environmental protection codes, building codes, safety codes, liability, animal welfare laws, etc. Consider that it is perfectly legal, and not at all uncommon, to kill a dog in China by inserting a blowtorch up its ass.

    Also, the laws in China are selectively enforced and it is possible for anyone with money to buy their way out of trouble. I consider China’s industrial polluters to be infinitesimally worse on a criminal scale than their murderers.

    But this study does suggest that Caucasoids are more likely to be criminals than Negroids, if it is legit and it is adjusted for race. So while it may not say anything against Neanderthals, it still doesn’t match the reality, unless we are to believe that most Africsn American felons have been convicted of crimes they did not commit.

    • Aeoli Pera says:

      With respect to neanderthals, what we’re saying is that significantly neanderthal-influenced features are abnormal by definition, and this paper describes detection of criminals as being basically the same as detecting deviance. Therefore neanderthaloid features will be flagged as predicting criminality because this is one type of “abnormal” face. Whether this particular type indicates criminality (predictive power), and then whether this indicates immorality (ethics vs. law), are different questions entirely.

  5. Edenist Whackjob says:

    Aeoli, do a post on System 3?

  6. Aeoli Pera says:

    >This needs to be the first entry for an Edenism Primer.

    Agreed. It’s an excellent proof of concept for phrenology.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s