Algorithmic Panopticon

Six screens in a circle, and you are standing in the middle.

Algorithmic Panopticon is a 22-metre-long video and sound installation developed with students, educators, and citizens. The work constructs an environment in which visitors are both observers and participants, embedded within a system that continuously generates and processes images of themselves.

At its centre, the installation examines how artificial intelligence is reshaping infrastructures of care and labour within education, healthcare, and midwifery. It frames these professions as sites where human judgement and machine-based systems increasingly intersect and negotiate authority.

Rather than presenting surveillance as a fixed condition, the work stages it as a dynamic process: a system of seeing, being seen, and being rendered visible through computational processes.

We built the installation around a six-screen cylindrical display running a single browser session across the full combined resolution. The browser connects over WebSocket to a Python backend that manages a queue of tasks and controls what is shown on each part of the installation at any given time.

The imagery is generated using Stable Diffusion running locally. Visitors initiate the work by photographing themselves on a connected device; that photograph is sent to the backend and passed through a sequence of image-to-image transformations at increasing diffusion strengths but with a fixed random seed. Each step using the previous output as its input. The resulting frames are assembled into a morphing video. A pre-generated artwork video plays continuously between interactions, with the backend timing the transition so new participant content is queued and displayed only when the previous artwork cycle has finished.