From images of wildfires and measles in the news, to images of the harms of smoking and alcohol in health education classes, visuals have long played a critical role in how we understand and communicate about public health. Until recently, if you saw one of these images or videos, you could take it as pretty solid evidence that something actually happened.
However, affordable and easy to use AI image generators have been rapidly changing how people relate to images. High-quality fake visuals can now be created by almost anyone and disseminated online through social media. Not only can these AI generated visuals lead people to believe false information, but they can also lead us to question the veracity of all the images and videos that we see.
In their recent article published in the American Journal of Public Health, a team of multidisciplinary scholars argues that the current environment may lead to what they term an “epistemic crisis,” where the erosion of trust in visuals significantly impacts the effectiveness of public health communication.
“How do we communicate about risk, crises, and outbreaks if visuals are no longer seen as trustworthy evidence about real conditions and events? And how to we protect people from visual misinformation created by bad actors? It’s a challenging balance, because you want to foster healthy skepticism without having people start to dismiss every image they see as AI generated,” noted Professor Linnea Laestadius from the University of Wisconsin-Milwaukee Zilber College of Public Health.
In their article, the authors offer a framework and ethical guideposts for how to begin to develop interventions and research studies addressing AI-generated visuals. They also highlight several recent cases where AI generated visuals have already caused harm, including deepfake doctors promoting scams on social media, fake images of California wildfires, and deepfake videos of recipients of Supplemental Nutrition Assistance Program (SNAP) benefits designed to amplify harmful stereotypes.
Katie Heley, Research Associate at Rutgers University and lead author on the study added, “This is a complex issue, but public health has the tools to make sense of it. As AI-generated visuals become more common in public communication, we need ways to map potential harms, to identify where to intervene, and to think through the ethics of how we respond, not just to react to what’s already happening. That’s the approach we take, pairing a structured framework for spotting risks with a guide for reasoning through the ethical questions.”
Interested in pursuing a PhD at the Zilber College of Public Health?
Students interested in working with Professor Linnea Laestadius, and earning a PhD in either Public Health: Community & Behavioral Health Promotion or Environmental Health Sciences are encouraged to connect via email at llaestad@uwm.edu.