2023/08/09 by Arsenii Alenichev, Patricia Kingori, Koen Peeters Grietens · 1 voice · 3 citations
Medicine · Social Sciences · #Misinformation and Its Impacts #Vaccine Coverage and Hesitancy #Viral Infections and Outbreaks Research
paper · pdf · doi:10.1016/s2214-109x(23)00329-7
openalex publication_date 2023/08/09 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/01
The Health Policy paper of Esmita Charani and colleagues,1Charani E Shariq S Cardoso Pinto AM et al.The use of imagery in global health: an analysis of infectious disease documents and a framework to guide practice.Lancet Glob Health. 2023; 11: e155-e164Summary Full Text Full Text PDF PubMed Scopus (0) Google Scholar showed how stereotypical global health tropes (such as the so-called suffering subject and white saviour) can be perpetuated through the images chosen to illustrate publications on global health. We propose that generative artificial intelligence (AI), which uses real images as a basis for learning, might further serve to show how deeply embedded the existing tropes and prejudices are within global health images. This in turn can perpetuate oversimplified social categorisations and power imbalances. Using the Midjourney Bot Version 5.1 (released in May, 2023), we attempted to invert these tropes and stereotypes by entering various image-generating prompts to create visuals for Black African doctors or traditional healers providing medicine, vaccines, or care to sick, White, and suffering children. However, despite the supposed enormous generative power of AI, it proved incapable of avoiding the perpetuation of existing inequality and prejudice. Although it could readily generate an image of a group of suffering White children or an image of Black African doctors (Figure 1, Figure 2), when we tried to merge the first two prompts, asking the AI to render Black African doctors providing care for White suffering children, in the over 300 images generated the recipients of care were, shockingly, always rendered Black (figure 3).Figure 2Prompt—a group of Black African doctors in the style of photojournalismView Large Image Figure ViewerDownload Hi-res image Download (PPT)Figure 3Prompt—African doctors administer vaccines to poor White children in the style of photojournalismView Large Image Figure ViewerDownload Hi-res image Download (PPT) Occasionally the renderings for Black African doctors presented White people, effectively reproducing the saviour trope that we were trying to challenge (figure 4). This was also the case for traditional African healers prompts that often showed White men in exotic clothing (figure 5), also posing the question of gendered biases in such AI-generated global health images.Figure 5Prompt—Traditional African healer is helping poor and sick White childrenView Large Image Figure ViewerDownload Hi-res image Download (PPT) Eventually we were able to invert only one stereotypical global health image, by asking the AI to generate an image of a traditional African healer healing a White child; however, the rendered White child is wearing clothing that could be understood as a caricature of broadly defined African clothing and bodily practices (figure 6). When requested to produce images of doctors helping children in Africa, the AI generated images of doctors and patients with exaggerated and culturally offensive African elements such as wildlife (figure 7). In further probing the rendering bias, we discovered that AI couples HIV status with Blackness. Nearly all rendered patients for an HIV patient receiving care prompt (150 of 152) were rendered Black (figure 8).Figure 8Prompt—an HIV patient receiving care, photojournalismView Large Image Figure ViewerDownload Hi-res image Download (PPT) In summary, we were unable to achieve our initial goal of inverting stereotypical global health images, and instead we unwittingly created hundreds of visuals representing white saviour and Black suffering tropes and gendered stereotypes; these images were created despite the AI developers' stated commitment to ensure non-abusive depictions of people, their cultures, and communities.2MidjourneyTerms of service.https://docs.midjourney.com/docs/terms-of-serviceDate: July 21, 2023Date accessed: July 21, 2023Google Scholar This case study suggests, yet again, that global health images should be understood as political agents,1Charani E Shariq S Cardoso Pinto AM et al.The use of imagery in global health: an analysis of infectious disease documents and a framework to guide practice.Lancet Glob Health. 2023; 11: e155-e164Summary Full Text Full Text PDF PubMed Scopus (0) Google Scholar, 3Alenichev A Encountering semiotic misdirection in COVID-19 etiquette guides.Sci Technol Stud. 2022; 35: 97-111Crossref Google Scholar and that racism, sexism, and coloniality are embedded social processes manifesting in everyday scenarios, including AI.4WHOEthics and governance of artificial intelligence for health: WHO guidance. World Health Organization, Geneva2021Google Scholar In unpacking this bias, it is essential to mention that AI learns by absorbing existing images available online, which, in the case of global health, have historically been considered as disrespectful and abusive. In contrast, AI is seen as a way of generating universal knowledge and products devoid of contexts and social meanings. This view reinforces AI-produced global health imagery as a so-called neutral visual genre, thereby misdirecting attention away from the context of the emergence of the original images,5Grietens KP Friesen P Gerrets R Kuhn G Kingori P Douglas-Jones R Misdirection in global health: creating the illusion of (im)possible alternatives in global health research and practice.Sci Technol Stud. 2022; 35: 2-12Crossref Google Scholar and the system through which they are now reproduced. Global health actors are already using AI for their media, reports, and promotional materials, making this an urgent, complex, and extremely relevant problem for science and society. A promotional marketing image in a photojournalistic style was shared on behalf of the WHO in May, 2023. The AI-generation detection tool Hive suggests a 99% likelihood that this image is AI-rendered, and the picture shows visible compositing artefacts and irregularities common among heavily edited or AI-generated images. This discovery leads to a sharp question for the international global health community: what is our responsibility in reinforcing or challenging visual stereotypes regarding socially enacted markers of similarity and difference, and the symbolic violence they might entail? And who is accountable for generating these visuals and potential transgressions? An extensive version of this case study was presented at the Oxford Global Health and Bioethics International Conference in June, 2023. Even though we helped create these problematic images by running prompts, and are displaying them for a wide diverse audience, we do not endorse the content of these biased AI-generated images, nor the counterparts from which AI draws from, and we hope to facilitate a conversation about the history, present, and future of global health and its visual culture. This research was funded in whole by the Wellcome Trust (grant numbers 221719 and 209830/Z/17/Z). We declare no competing interests.