Author: Niru Varma
Artists: Nirvan Marathe and Rahel Kiss
Editor: Alexia Correia Ferreira
Picture a wanted poster for a dangerous criminal. Emblazoned with a picture of their face and a description of the crime committed. Now imagine that the face on the poster isn’t really the face of the suspect at all. Rather, it’s a probability-based estimation of what their face might look like according to their DNA. That describes the image tweeted out by the Edmonton Police Department in Alberta, Canada. They had used DNA phenotyping, a way of deducing an individual’s appearance from their DNA to assign this face to the perpetrator of a violent sexual assault.
The forensic DNA phenotyping (FDP) apparently determined the suspect to be a Black male (as previously described by the survivor to the police) of East African descent with brown or dark brown skin, brown or black eyes, and black hair. Alongside this information was a portrait, a computer-generated image of a young Black man, claimed to be an approximation of the suspect’s appearance based on a DNA sample.
The big issue? There is absolutely no evidence to suggest that DNA phenotyping can be used to predict individual faces. As an extremely complex trait, it is likely that facial morphology is determined by a large number of DNA variants, each with a small effect, making them difficult to identify and therefore mostly unknown. Parabon Nanolabs, the company behind the Snapshot commercial phenotyping service, has never put out peer-reviewed publications validating their methods nor disclosed what markers they use for establishing facial characteristics, casting doubt on whether there is any validity to their methods at all. While their predictions of skin colour, eye colour, hair colour, and biogeographic profile (often simplified as denoting ethnicity) are based on established methods and markers with evidence to support their accuracy, these are simply predictions of a range of phenotypes based on probability as opposed to a definitive description of the suspect. Additionally, unlike a typical facial composite composed from eyewitness testimonies, an FDP image cannot depict distinguishing features like scars, blemishes, facial hair, hair style, or tanning.
All of this means that the image produced is very generic, bearing a possible resemblance to a large number of individuals, whilst not even necessarily representing the suspect with any degree of accuracy. An informal test by the New York Times showed how many individuals misidentified FDP images of their coworkers, with one image having a 0% correct identification rate. This perfectly demonstrates the potential consequences of publishing these images claiming them to be the faces of dangerous criminals. The labelling of the image as an “approximation” does not necessarily prevent scrutiny and even harm towards individuals who may resemble it — after all, the purpose of the image is to incite suspicion towards those who fit the description. If those viewing the image are truly not supposed to treat it as an accurate image of a potentially dangerous criminal, then what is the point of it?
Even outside of the questionable use of FDP to predict faces, its use in determining the race and skin, eye, and hair colours can also be called into question. Some would say that instead of identifying a specific suspect, they instead create a suspect population. This, of course, is nothing new; even before the advent of DNA technology, security cameras and eyewitness accounts could testify that the suspect was, for example, a white man with brown hair — but FDP is unique for being unable to show any other identifying characteristics and instead flattening suspects into just a race and colour profile. In the Edmonton case, the FDP did not reveal any new solid information that could help to catch the suspect given that the survivor had already identified the suspect as a Black male. In a case where even less is known about the suspect, it could be argued that creating a suspect profile by means of FDP results in police treating a whole demographic as potential perpetrators as opposed to helping them to find a singular perpetrator, increasing the risk of harm towards that demographic.
While the Edmonton Police apologised and took down the image as a result of a spike of public backlash, claiming to have only used it as a last resort, they are far from the first police department to use this tool and will likely not be the last. Parabon Nanolabs point to the customer satisfaction from police departments using their service as evidence of its effectiveness. With their appeal to public safety and to the comforting notion of having dangerous criminals taken off the streets, many will side with FDP providers and the police that use their services, believing FDP keeps them safer. This is bolstered by the fact that FDP is (supposedly) based in science and data — this, in the eyes of some, makes it automatically objective and free from human biases and error. A composite sketch may not be accurate because human memory isn’t always accurate — the idea of an objective image based on cold hard evidence is an appealing one. Lack of understanding of and misplaced belief in science can, in some cases, be as damaging as denial of it. It’s for this reason that those in the field of genetics have a responsibility to ensure that knowledge of this highly socially relevant technology does not remain in the ivory tower and is spread as widely as possible, so that the public and lawmakers can be better informed about genetic technologies like FDP, and their potentially dangerous consequences on society.
kinesismagazine.com 