Walter Benjamin’s well-known piece the “Work of Art in the Age of Mechanical Reproduction” has long been a canonical essay on the role art plays in the age of automation. Benjamin saw art both as fueled and altered by mechanization. In a conversation with artist Heather Dewey-Hagborg, a partial transcript of which follows below, the role of art in the age of digital reproduction, to paraphrase Benjamin, emerged as a critical theme. Heather’s work, which spans over a decade, is a complex meditation on the contemporary experience of widening digitization. Her work Stranger Visions is perhaps the best known: a project where she reconstructs faces from DNA left on refuse she’s found on the street – a chewed up piece of gum, a stray piece of hair, a lip stain on a glass – into voluptuous, three dimensional portraits. During our conversation, we talked about the creative and the intellectual process that has shaped Stranger Visions, but also Heather’s more recent portraits of Chelsea Manning and her latest project Invisible.
Heather’s work invites to see notions of art and artistry, of craft and craftiness, as a way not only of evaluating and categorizing her relationship to data, but as a way of thinking through that which contours data. Heather’s process of working with data, of surfacing faces from the data she discovers, materializes and performs data as a form of social construction, as itself an art. But it likewise invites questions. Where art is often seen as a site of potentiality, as that which can help to subvert taken for granted categories, it also operates, particularly in ever-increasing distinctions between data art and data visualizations, as that which compromises the ability to take the art and artistry that produces data at face value.
Yuliya Grinberg: Of course, I want to ask you a little bit about your project Stranger Visions. Can you give a brief summary of how that project began?
Heather Dewey-Hagborg: I have been working with the themes of surveillance and machine learning in my art work for, like, 10 years, if not more, at that point. I was mostly thinking about it from the electronics side of things, thinking about facial recognition, and speech recognition, and the Patriot Act. And cameras on the street. And the warrantless wiretapping. And working with the algorithms that would be processing this data. And also thinking about the emergence of Google. And the consumption and the commodification of our data. And I just had kind of a revelatory moment where I was sitting in a doctor’s office and saw this hair sticking out of a crack covering a glass print. And I became totally transfixed by this hair thinking about whose it could be and what I could learn about someone from their hair and really relating to this person and thinking of all of the times I had my own hair caught in one of these things — especially in NYC — these contraptions that hold the posters. I would always get my hair caught in one of these on the subway. So I started to really relate to this person and thinking about them.
And then when I left I really began wondering how much I could learn about someone from something like that, from a hair that had been accidently left behind. I started noticing all of these cigarette butts, and chewed up chewing gum, saliva on a rim of a cup much like I am sipping on now – things we have been leaving around all of the time without giving them a second thought. And it just occurred to me that there was a huge surveillance risk that no one was talking about. At least not since Gattaca in 1997. And I started researching this and trying to find out what could I really learn about someone from some little bit about themselves that they left behind, from a forensic artifact. And the conclusion I came to was: it’s complicated, but really it’s quite a lot. Scientists were publishing quite a bit on the one hand about what DNA might say about who we are and also specifically working on extracting these things from forensic traces. And so I discovered that there were already several scientists who were publishing in this field called forensic phenotyping. So attempting to create an image of a person from nothing more than their DNA. I started pulling this research together and speculating even further where I imagined this research could go beyond descriptions. For example, quite a few scientists were working on things like hair color and eye color.
And I started thinking well, where might that go, and imagining that that might go towards the form of a portrait. And I started looking into all of the other genetic traits in a very preliminary way that might influence the construction of a portrait. So then in Stranger Visions, I pulled all of that research together and wrote code that would do that. So basically that would move from the sequence data to a speculative portrait of what a person might look like based on their genomic information. And then I went around the city and collected different artifacts like chewed up gum and strands of hairs. I think a lot of it was, well maybe that could be possible. Let’s see. And so I put those things together to produce the portraits that you see in Stranger Visions which are full color, life size 3D prints representing these strangers from their DNA.
Y: Your project is very interesting and it invites the idea that in creating an image you just copy, paste, and print out. It’s very much part of our computational imaginary.
H: Yes, as though you go directly from DNA code to output face. Yes. Definitely true.
Y: And it seems to speak directly to this idea of DNA as code, or data more broadly as a copy, as our digital double, as we hear it said all of the time. But your project pivoted from a straightforward, biological approach, from, ‘let’s see how much data can tell us about ourselves,’ to ‘what what kind of cultural assumptions are encoded in that data to begin with?” How did you come to this shift?
H: So I think the project is, in a way, forever a work in progress. Or at least it is work in progress that spans many years for me, even in terms of my own understanding. So when I began working on the project, I had no idea what I was going to find and I had some sense of problems around genetic determinism. So I was always peripherally aware that this was a problem. My description of the project was to call attention to the impulse towards genetic determinism. But the real details of that and the politics of that became progressively more apparent to me in the years of working on it.
So when I started, it was really mainly about surveillance, about this kind of revelation about the vulnerability of the body, and about this wish that we would be able to create a face from our genetic code. That we could think of DNA as a code even, as this book of life.
But as I worked on it and as read more and more, and really as I enacted the process – I constructed the algorithm myself – from reading scientific papers and so forth, the problematic became increasingly more apparent to me.
It’s interesting in a way that the act of doing something really brings you into a different level of acquaintance with the problematics of the material. If I hadn’t walked through it myself I wouldn’t understand why it was problematic to create portraits based on ancestry assumptions. Really, how much that is the imposition of a stereotype for example.
I don’t think that the project would have been nearly as successful, on many levels. Not just successful in terms of getting attention, but also as a work of art –if I had not adopted that problematic practice and shown, sort of embodied, this desire for deterministic code by creating this singular portrait, that is associated with one person’s DNA, using the computation process. All of the bits of it mattered.
Y: Can you walk me through one example? You talked quite a bit about racist assumptions encoded in our DNA, can you walk me through the kinds of things you had to engage with as an artist to get to a face that appears to simply emerge from the data?
H: Yeah, and in addition to the racial stereotypes I would also throw gender stereotypes in there. Which is something that I point to in the portrait of Chelsea Manning as a follow-up project. But the artistic choices are really enormous.
Basically the program that I wrote would generate a version of face. Imagine you feed in these characteristics. So for myself it would be pale skin, and freckles, and blond hair, and less tendency to be overweight. And things of that nature. And then imagine all of the possible faces you could have from that. I mean, there are so many! And so what the code would do is generate random permutations based on that data, based on that list of descriptors. I would generate five different versions of a face, and then choose the one that I thought was the most interesting and compelling as a sculpture.
Y: What were the things that you would find yourself looking for? What makes an interesting face?
H: I mean, it’s just something that spoke to me. I think either I would see a face and relate to something that felt plausible as a real face. Certainly nothing that would look too caricature-like. Although, the way that I wrote the code generally would reign things in towards the realm of the plausible. Sometimes if they looked like someone that I knew or someone that I’d seen. A lot of those samples were collected in my neighborhood and in particular this piece of gum that I collected outside of a bodega by my apartment. The pieces that I generated from that, to me, looked just like the guy that was working at the Radio Shack down the block. So I chose that one because it looked so familiar.
Y: Yeah, I mean, looking for the familiar in a face could be really problematic and complicated.
H: Yeah, we do that even in regular wanted posters and things like that. Oh – that face looks familiar. People look a lot alike in a lot of ways.
Y: I wanted to ask you about the portrait you made of Chelsea Manning. That was super interesting because you made one abstracting gender and another one you sequenced for gender. What other decisions went into that process?
H: When I was contacted and asked to produce a portrait of Chelsea Manning it seemed like a good opportunity to call attention to the limitations of the technology. So a lot of thinking I did around the piece went into that. So that’s how I settled into the idea of doing a diptych. So producing these two different portraits that would represent her – one that would be gender neutral and one would be gendered “female.”
Y: And you left male out.
H: Yes, and I left the male out. What I decided to do was not to profile genetic sex at all. Normally that would be the first thing, the first thing in any kind of forensics would be to look at sex. So I just didn’t look at that at all. And the first portrait that I produced was not gendered. So it shows a kind of neutral face of this algorithmic model. And the second was was gendered female, with which she self-identified. So I decided those would be the two that I would show.
Y: The neutral version, what other parameters did it draw upon to generate a face?
H: The way these face models are constructed is that lots of different people’s faces are 3D scanned. And then the whole space – so if you imagine that you have a multi-dimensional face that allows you to move between the differences in people’s faces, then you can also organize a face along the many different components in people’s faces. One of the principal components is associated with sex. So it’s almost as if you have a volume slider that you are moving between male and female. And that is how it is presented within the computational model. It’s baked into the model, I would say, through the construction of the data, through the choosing of exemplars, through the choosing of the data … all of that factors into how sex/gender is constructed within that computation space. And so the computation models that I appropriated already had this ability to sort of morph male/female. So it would be like, it would be like you put in a value of – 5 it would be female and +5 it would be male, those are the gender extremes. And then 0 would be supposedly a neutral face. And basically what those represented were specific traits. So I would say, like, a wide jaw was associated with male, and a thick brow, and things like that.
And then I generated lots and lots of different versions of the face, again. And I chose the one that was really compelling and spoke to me and felt like it could be Chelsea.
Y: Have you seen photographs of her since starting the project?
H: Well, not since … The only photographs that there really are of her are as Bradley Manning prior to the transition, so I’ve seen the same photographs that others have seen online.
And then there was a portrait sketch done of her –a drawing. She worked with a portrait artist. So I’ve seen that as well. So sort of like, how she sees herself. And so, of course there are factors in choosing the faces, no question. It is totally subjective.
Y: You know, we started this conversation with this metaphor of the copy and of printing that in a sense is already embedded in the project, but the more we talk about it, the more I hear you talk about the project, sculpting has emerged as a more operative metaphor. We often talk about “massaging data” – that phrase gives that same sense of information being something that is pliable rather than absolute. Part of it is the ability to manipulate things, but it also in a productive sense, it’s a creative ability.
H: Absolutely, you can tell so many different stories, you can create so many different faces from the same body of data. So for instance Chelsea and I have almost the same set of traits. I have more of a tendency for freckling – otherwise there are really no big differences in our DNA. The same kind of ancestry. And blue eyes. Skinny.
I could have almost just put my self-portrait on display.
There are so many ways in which that data could be shaped.
Y: You deliberately call your work “art.” What do you think “art” offers in your work? What kind of space does it open up or foreclose?
H: For me, speaking from the voice of art is the honest voice of my process and what I am doing. I would not put something out there other than art. But I think probably a lot of things that are put out that want to have an authoritative voice are not as authoritative as they may present themselves. Probably the things that we consider art should be a much bigger category.
Y: What I thought was really fascinating was when we spoke about your project Invisible being classified as art by the New Museum store. What does that label say about that form of expression
H: I wish I would have taken a riskier route. I sort of played it safe by allowing Invisible to be labeled as art.
I didn’t start as art, actually. When I launched the project and when I did initial round of interviews around it, I wasn’t labeling it as art at all.
Y: How were you calling it?
H: I just said that I started this new Biotech company. I presented it that way. I launched it at this industry conference called the Bio-IT World Expo. I got up on stage and said, ‘We need some way to protect our genetic privacy, basically.” And I played the commercial. And the plan was to really release it as a product. But I wasn’t sure what channel to do it through.
But then this new opportunity arose to show it in the New Museum store. And I was a little worried because I had some legal questions raised about Stranger Visions. And I was just starting a new faculty position. And I really was worried that I would get myself into some hot water. Several people advised me in that capacity not to push it too far.
Y: So what kind of permission does Invisible as a product have that Invisible as art does not have?
H: I think that when it’s art, unfortunately people think that it’s not real. It seems like design fiction. It seems like speculative design. And it takes away some of the danger. But really the whole point of the project was to point to some of the danger.
Y: What was the response like at the Expo?
H: Oh, it was really good. It was taken seriously, more or less. And people asked real questions about it as a product. And asked things like would I give a key to the police that would give the genetic fingerprint so they would be able to see if it had been used. It was really good. So it just goes to show that if there are so many ways these things could go, the decisions that we make are really significant. It could have gone so many different ways, just by deciding where I would sell it.