Deep Dream Vagine Machine – Parallels between Sexual Consent and Consent for Data Usage

This is going to be a relatively intimate post for this blog.  It does, however, involve data & machine learning in different forms as well as consent and the ethics of the data used and the application of these concepts brought down to ground level in the form of an art project that, over time, begged questions about the collection and use of people’s private data, contextual integrity, and the need for ongoing conversation of consent as context changes.

A few years ago, I became aware of a neural network known as ‘deep style transfer’ that allowed a user to transform an image using the artistic style from another.

I was consumed by the infinite combinations of stylizing everything, but realized that transforming my own selfies was not a sustainable hobby. I needed more content.  Something unexpected, challenging to obtain, with texture and variety.

To make this fun, I crowdsourced from Facebook friends.   “If anyone wants to send me vagina pics (your own vagina only), I will do something fun with it,”  I posted. Soon after this invitation, I received my first anonymous vagina photo. I did not need to know the identity of the women, so long as they were consenting adults (my audience for the post was a limited portion of my friend list, and I don’t have any friends under 18).

So began my year long journey of making stylized vagina portraits for friends, a project now titled “Dream Vagine.”  I sent them the results and requested consent to post the images anonymously on my Instagram. As I received support, people asked if I would show them in a gallery, sell them, and maybe make lots of money!  To avoid getting in the weeds about my philosophy on art, I will jump straight to my discomfort of how I would hypothetically be handling the privacy of these women if I was to take up that offer. It would not feel right to me to take these women’s images and post them in a gallery with the intention of profiting.   Since my photo was also part of the project, it was easy to put myself into the shoes of any other women. To put this dryly into Mulligan’s framework of the dimensions of privacy:

  • Dimensions of theory
    • object: dignity, control over personal information
    • justification: to protect from social or professional consequence as well as undesired use of image
    • exemplar: sexual shaming, or unwanted use of private photos
  • Dimensions of protection
    • target: body images
    • subject: women in the photos
  • Dimensions of harm
    • action: use or distribution of the photo without consent of the subject
    • offender: me or anyone on the internet
    • from-whom: everyone, employers, family, friends
  • Dimensions of provision
    • mechanism: communication, social norms
    • provider: me
  • Dimensions of scope
    • social boundaries: Instagram, anywhere the photos are stored or displayed
    • temporal scale: indefinitely until otherwise communicated
    • quantitative scope: per-case

To summarize, a woman whose photo was submitted to me for this project would want to protect herself from the use of the photo in an undesired manner, by me or by other people.   The act of publishing the stylized photos was a way to exhibit woman’s photo under consensual circumstances in contrast to the stories we often hear of people getting hacked and having their nude photos exposed without their consent.  Even though this project was a form of protest against the cliche outcome of having a nude photo shared without consent, technically, there is still risk of misuse within the project. I wanted to avoid that, but how would I know if I was misusing the images?  

It was important to consider the context in which the images were used or displayed.  When the contextual integrity of data is static, the worry about loss of control is minimized. The image is used in one place, for one purpose, at one time, for a set audience.  The change of context is precisely what many of the women worried about. In this project, the context changed every time the image was displayed. So, if I wasn’t going to maintain contextual integrity, how could I maintain the comfort of the women?  How would I know what they were comfortable with? This was possible by maintaining communication. The only way to know was to ask each woman. If I continued to communicate with them about how their image would be used, I could obtain consent again, at every layer of context, as it changed. This allowed me to progress the project to the point that it could be displayed at a gallery.  Further, if a viewer wanted to request a print, they could submit a request that would be relayed to the woman who would then consent to the purchase of her image by each individual.

As this story unfolded, it paralleled the widespread sharing of sexual assault stories through social media.  The vagina portraits became more to lure viewers in to discuss the topic of consent in a sexual context in a different way.  Sexual communication could be fun and not just some kind of legal obstacle. The request for consent is, actually, the coupled expression of desire and respect. “I want you, but more importantly, I respect you.”

This particular sexual consent conversation stacked perfectly on to the consent conversation for use of the visual information in these private photos.  The portraits themselves were no longer the focus of the art, they were simply a by-product and a manifestation of the wonderful and unique things that can come from the consent of a woman for use of her body, and use of her information.

Leave a Reply