Artificial Images in Global Health

AI does not exist in a vacuum. The earlier, pre-AI ‘quasi-real’ images are being fed into current algorithms to produce similar synthetic images on demand, allowing for ‘layers’ of fakery to unfold.

The use of such synthetic images is —according to its supporters—justified as it avoids identifying and potentially harming affected individuals. Using AI, it is said, reduces the burden on communities that are over-represented in images of human suffering, and it circumvents issues of exploitation and danger. From a pragmatic marketing perspective, mobilising AI reduces labour costs and enables tighter control over visual messaging.

However, it still often draws from real problematic images, reproduced de-novo, albeit synthetically. The history of humanitarian photography is peppered with examples of image manufacturing and tinkering associated with, on the one hand, the limitations of nineteenth century equipment, and on the other hand, feeding into morbid curiosities and manipulating emotional responses. The practice of fakery has persisted despite twentieth century photography advancements that allow for candid, in-situ pictures. Photo-editing computer software made more accessible the techniques and tricks previously limited to the darkroom.

Beyond those techniques, photographers on assignment for agencies or organizations might rely on posing community members to reenact their—or imagined others’—realities. While staging, reenacting and repurposing of images has been a regular practice in global health photography since the early days of the medium, generative AI puts that creative potential in the hands of anyone with access to the technology. No human connection required.