It sounds like the setup of a Philip K. Dick story. In a new study recently published two Chinese researchers, Xiaolin Wu and Xi Zhang, from Shanghai Jiao Tong University say they have developed an artificial intelligence that can tell if someone is a convicted criminal just by studying their facial features.
The AI looked at 1,856 mugshots, nearly half of whom were convicted crims, and the program was able to pick out various facial features and guess correctly from those 90% of the time which of the people were criminals. The "discriminating structural features for predicting criminality" were "lip curvature, eye inner corner distance, and the so-called nose-mouth angle."
Although the researchers say that "Above all, the most important discovery of this research is that criminal and non-criminal face images populate two quite distinctive manifolds." the idea that you can tell criminals from non-criminals just be looking at their face is, quite rightfully, thought of as unethical and unscientific.
The idea that facial or any physical characteristics correlate to someone's personality has long been thought of as complete nonsense. This new study seems to be the latest iteration of a centuries' old pursuit to label people as criminals by the way they look.
And, as such, has caused quite a bit of backlash. Which means it's unlikely that this will ever be used for actual policing, not by the US anyway.
"This is no different than Craniometry from the 1800s, which has been debunked." Dr Richard Tynan, technologist at Privacy International, told The Telegraph. "In fact, the problem runs much deeper because it can be impossible to know why a machine has made a certain decision about you.
"It demonstrates the arbitrary and absurd correlations that algorithms, AI, and machine learning can find in tiny datasets. This is not the fault of these technologies but rather the danger of applying complex systems in inappropriate contexts."
China has already been testing pre-crime technology with its "predictive policing AI"—hopefully they won't combine that with the results from this study.