The dystopian future is one step closer to reality, thanks to a new computer algorithm developed by a pair of scientists from Shanghai Jiao Tong University that is supposedly able to distinguish criminals from law-abiding citizens just by analyzing their faces. While some may take this news as the biggest indication yet that we are all living in a Philip K Dick story, it’s unlikely that this technology will ever actually be used to preemptively arrest people for their future crimes.
After observing 1,856 mugshots, roughly half of which belonged to convicted criminals, the computer was able to identify a number of distinctive characteristics that allowed it to reliably distinguish outlaws from “inlaws”. Thereafter, it was able to predict with almost 90 percent accuracy which faces belonged to transgressors and which did not.
In their paper – currently available via the preprint server arXiv – the researchers say that their algorithm found key differences between the two groups, including “lip curvature, eye inner corner distance, and the so-called nose-mouth angle.”
Top row shows the four criminal face types; bottom row shows non-criminal types. Xiaolin Wu, Xi Zhang
They also say that criminal and non-criminal face types can be more generally categorized as belonging to one of two main “manifolds”. Three different facial structures were found to belong to the law-abiding manifold, while four were identified within the criminal manifold.
“In other words,” write the study authors, “the faces of general law-biding public have a greater degree of resemblance compared with the faces of criminals, or criminals have a higher degree of dissimilarity in facial appearance than normal people.”
Unsurprisingly, the study has stirred up a considerable amount of controversy, with many commentators pointing to the fact that not only is this type of categorizing completely unethical, but it also falls into a distinctly grey area of scientific reliability, as all associations between physical characteristics and personality traits have been well and truly debunked.