Revell, Timothy Concerns as face recognition tech used to ‘identify’ criminals. New Scientist. December 1st 2016.
Garvie, Clare, Bedoya, Alvaro and Frankle, Jonathan The perpetual line-up: unregulated police face recognition in America. Center on Privacy & Technology at Georgetown Law. OCTOBER 18th 2016.
Is there really a criminal face? I don’t think the research discussed in the New Scientist article settles the debate by any means, but at least the controversial idea is opened up for investigation. If there is one my guess is that it is a look that coincides with the Australian face (every race and nation has a distinctive averaged facial type, apparently). European colonisation of Australia began as a penal colony and thus a good part of the white genetics of Australians arrived in this country through people identified as criminals. My best guess is that the crim face has a large straight nose, thin lips and puffy, small eyes. I’d guess this unattractive face could in itself be a social and economic disadvantage, or could be symptomatic of a phenotype that includes some degree of intellectual impairment. I think if there is a crim face it might have little to do with personality but a lot to do with disadvantage, but this is all speculation.
I think it is worth noting that claims made in the print version of this article about supposed advantages of AI over humans in face recognition skills such as identifying age, gender, ethnicity and tiredness by looking at faces presumably only apply to humans of average face recognition ability who maybe are not as exhaustively trained in these skills as the AI systems have been. One cannot compare human ability with AI in face recognition until appropriately trained super-recognizers (representing the top end of human ability) have been pitted against machines. I’m guessing this hasn’t been done.
Perhaps the most important part of this article is right at the end; “…the majority of US police departments using face recognition do little to ensure that the software is accurate.” That certainly is not good enough. Human super-recognizers have abilities that have been proven in scientific testing and also in practice in policing in the UK. Why do so many people persist in the assumption that machines must be better than humans in visual processing, in the face of an abundance of evidence? The link in the New Scientist article to the website of the researchers who have criticized the use of face recognition technology in law enforcement in the United States of America is worth a look for sure.