From evidence to testimony: how AI-generated images of refugees can make demands

The essay looks closely at Exhibit A-i: The Refugee Account (2023), a collaborative human rights project communicating the experience of 32 refugees detained by the Australian government in offshore processing centers under inhumane conditions. To make up for the lack of images, they used the AI-powered text-image generator Midjourney. Analyzing the project, I am interested in how to conceptually approach these images that refer to the history and theory of photography but do not share photography’s ontology. Drawing both on traditional photo theory and memory studies as well as literature dealing with algorithmic images, I make an argument for these images to constitute a new kind of testimony, where individual memory and collective image archive intersect. To do so, I draw on select modes of production of AI-generated images, detailing aspects such as the linking of text and image as well as the reliance on noise to make a representation. While acknowledging the power asymmetries between the Global North and Global South inherent to AI, I argue that projects such as Exhibit A-i can appropriate the technology to question its use in migration policies, all the while demanding from the audience to reflect on its own implication.