The Stranger Could Be You: Artist Heather Dewey-Hagborg on Her Chelsea Manning DNA Portrait
If you're doing a story on someone you can't photograph -- say, because she's serving a 35-year sentence in a military prison -- your best bet might be contacting artist Heather Dewey-Hagborg. As we put together our September issue's feature on whistle-blower Chelsea Manning, we recalled Dewey-Hagborg's 2012-2013 project Stranger Visions, a series of 3D-printed portraits created using DNA harvested from cigarette butts, spat-out gum and other street detritus. Next thing we knew, a FedEx package containing cheek swabs and hair trimmings was making its way from Fort Leavenworth to the Chicago-based artist, who then extracted, processed and interpreted Manning's DNA; the result is the uncanny portrait, or "forensic DNA phenotype," above. Of course, another challenge was that Manning, who announced her true gender identification after she was sentenced, as Bradley Manning, in 2013, is virtually invisible: a black-and-white photo from 2010 and an illustration on her Twitter account are our only references. We came to Dewey-Hagborg with this dilemma and were thrilled to learn that she could control the gender parameter in her DNA-generated portraits, either by inputting it manually or leaving it out altogether. (The version we chose represents the second option; you can see the first below.)
Once we'd closed the issue, we Skyped Dewey-Hagborg to learn more about her process and the surprising parallels between Manning's work and her own.
Generally, how much control do you have over the data that goes into these portraits?
A tremendous amount. It's a very artistic practice. It's much like if you were generating a police sketch by hand -- it's that level of control, in the sense that you can parameterize gender along a spectrum. There's a continuous variable, so you can set it to more male, less male, "neutral," more female, less female.... But it's important to remember that this kind of model -- called a morphable model -- is based off of 3D scans of people's actual faces, so the model is a generalization from the original faces that were input. Depending on whose faces those are, different kinds of stereotypes of what gender looks like become embodied by that model. So there's only one way of looking female in the model, and that's one of the limitations. It's one of the things that I've been trying to highlight in my critique: that things like gender and race become problematic because they rely on very simplistic kinds of stereotypes.
How much of this software is your own design, and how much are you inheriting technology that has these restrictions?
Basically, I took a morphable model that had already been developed by an academic research unit and started exploring it and sort of hacking into it. Then I added this front end that takes in various kinds of DNA data and creates a phenotypical profile from that, and that's fed into the parameters of the 3D morphable model. One of the major problems is that the model I worked with was developed in Basel, Switzerland, and that's a very, very white place. It was really challenging to try to change the model so that it could actually represent people who weren't white. I basically ended up taking their model as a framework and then retraining it with different kinds of images. I think that those kinds of biases are always built into these data sets.
Walk me through the choices you made while creating this portrait.
I got the FedEx envelope, and I wasn't sure what I was going to find when I opened it up, and so it was particularly exciting to see the plastic bag with her return address label on it. Inside the bag was a bundle of hair clippings and cheek swabs. I did a DNA extraction on the samples and quantified the DNA I extracted. Then I started going through a series of polymerase chain reaction experiments where basically I'll amplify a small section of DNA and look at a specific trait that's associated with that section. Normally I would be looking at a gene that's on the Y chromosome to determine if the person basically is male or female, and in this case I decided to skip that step. And then I looked at the genes for eye color and ancestry and tendency to be overweight or not... Each of those is basically an independent experiment: you have these little tubes and you put in the DNA and you put in these things called primers that are kind of like bookends that mark which part of the genome you want to amplify and examine. I'll run it in a machine and take the product of that and run it through gel electrophoresis, which will allow me to see if the amplification was successful. Each of those reactions gets FedExed to a company that will take that product and generate a sequence file from it. And then I take those sequences and bring them into bioinformatics programs and align them with known reference sequences, from the Human Genome Project, for example. I'll take the output of that and feed it into the front end that I developed for the morphable model. All of the variations -- what are called SNPS, or single nucleotide polymorphisms -- get entered as data into the front-end program, and the front-end program will generate a version of a face based on that information. Then there's this process of generating and choosing: I'll generate lots of different faces -- different versions of this identity -- and I'll go through and decide which one I think is the most compelling.
So what were your criteria for finding the most compelling picture of Chelsea?
Obviously, since I already know what she looks like, that does very much influence my choice. Normally, if I was just working on a stranger's DNA, I would simply be looking for a face that I thought was interesting, but in this case I definitely was leaning more toward ones that I thought looked the most similar to Chelsea -- but not necessarily looking like her in the photos that you see blasted across the Internet. It's my interpretation, or my guesswork, of how she would want to be represented.
Because we're only talking about one widely circulated photograph of Chelsea, and one approved illustration -- that's all we've got. Were you looking at photos of Bradley and using your imagination as to what Chelsea might look like?
Yeah, exactly. Because the two pictures of Chelsea are highly dominated by hair, and as you know the portraits I generate don't have hair, so there was a lot left to the imagination. I just tried to put myself in her shoes and try to imagine which of the faces would be the most representational of her identity.
A rendering with the gender manually input as female
We also tried a round where you input the gender as female. What were your thoughts on the two versions?
I think it's interesting to see them side-by-side. That would be a fascinating exhibit -- to see any person's face parameterized along that spectrum, because it really does call attention to this stereotyping of gender and this continuous variable of it, and how subjective these kinds of judgments are. It also raises the question: What is a female face? Do we need to change someone's face to make them look more stereotypically female to consider them female? I would say no. I think that the more neutral face makes a stronger statement.
What parallels do you see between Chelsea's work and yours?
I think of a lot my projects as "exploits," in the hacker sense of the word: taking a vulnerability and calling attention to it. I've been thinking about whistle-blowing a lot because one of the real capacities of art, in terms of challenging oppression, is a power of revealing or truth-telling, in the sense of the Greek cynic: someone who would speak truth to power. In my own work, I see that as twofold. One is the exploit of showing this vulnerability that we are shedding DNA all the time and it's increasingly easy to take that and find out all kinds of things about us. And I would say that Chelsea also does this: she shows us the vulnerability of information. I also think in terms of what is actually revealed. For me, in Stranger Visions, it's taking something that was behind the scenes at the laboratory and making it as public as possible, to try to get it into public consciousness and get a dialogue going about it. I would never have the hubris to say that [my work] was remotely as significant as Chelsea's, but I think the parallel would be what the artwork shows. I think often what visual art does is it shows rather than tells that something is possible. In the case of whistle-blowing and releasing documents, there's the stage of sifting through all those documents and deciding what gets done with it; in the case of phenotyping, this was something I wanted people to have a visceral experience of. I wanted them to look it in the face -- to have that face-to-face encounter with a stranger who could be them.