Overview

Created for the Debox pavilion/exhibition hosted by Public Access Miories as part of The Wrong Biennale, Consensus Sensing is a speculative design and data exploration experience that is a continuation of my series of investigative experiments into the transphobia baked into AI image general. This project expands my image series Consensus Portraits which is a series of portraits of trans and non-binary people, generated with popular generative AI tools. It has long been known that biases contained within datasets surface in generative “AI” systis trained on this data. The generated images in this series highlight the transphobic tropes baked into so-called AI systems and contain subtle tells about the source imagery used in their creation. For Consensus Sensing, I have chosen to focus on images made with the prompt of generating a transwoman.

Consensus Sensing presents this archive of AI images within an online custom user interface containing footnotes, context, and overlays, inviting viewers to closely read the images. The user interface is a machine from an imagined open hardware future, positioning the viewer in a future looking back at this archive. This positionality aims to both imagine a more open and humanist tech future and also to give the viewer a sense of distance from the generated images. Invested in looking critically at the data and process obscured by the black box, Consensus Sensing offers the audience an opportunity to slow down their viewing and glimpse the inside of the black box.

The project can be viewed permanently here:

https://cthompto.github.io/consensus-sensing/

The source code for the project can be found here:

https://github.com/cthompto/consensus-sensing