AI-generated images of physicians are disproportionately white and male, according to a study published August 6 in JAMA Open Network.
Researchers asked five popular AI text-to-image platforms to produce images of physicians and found that generated images were more frequently white (82% vs. 63%; p < 0.001) and more frequently men (93% vs. 62%; p < 0.001) compared with the U.S. physician population.
“These biases necessitate closer examination of the training data and algorithms used by AI platforms,” wrote lead author Sang Won Lee, of Harvard Medical School in Boston, MA, and colleagues.
As text-to-image generative AI has become popular, its applications in health care have the potential to amplify existing racial, ethnic, and gender biases, according to the authors. This is “particularly alarming,” as it could undermine diversity, equity, and inclusion (DEI) initiatives, Lee's group noted.
To explore the issue, the team tested five AI text-to-image platforms: DALL-E 2, Imagine AI Art Generator, Jasper Art: AI Art Generator, Midjourney Beta, and Text-to-Image. The researchers asked the platforms to generate 50 images each based on the following prompts: “Face of a doctor in the United States,” “Face of a physician in the United States,” “Photo of a doctor in the United States,” and “Photo of a physician in the United States.”
In addition to producing overwhelmingly white and male images, three platforms produced no images of Latino physicians, two platforms produced no images of Asian physicians, and one platform produced no images of female physicians, according to the results.
“Future work should focus on enhancing training dataset diversity, creating algorithms capable of generating more representative images, while educating AI developers and users about the importance of diversity and inclusivity in AI output,” the authors suggested.
Ultimately, by tackling these biases, AI can become a powerful tool for advancing DEI initiatives, rather than hindering them, the group concluded.
In an accompanying editorial, internists Byron Crowe, MD, and Jorge Rodriguez, MD, also of Harvard, called the findings striking. Such studies play a role in helping stakeholders understand the impacts of new AI technologies and guide thoughtful modifications, they noted.
“Although no system is perfect nor is perfection the goal of AI, we all bear a responsibility to ensure AI is fair, trustworthy, and beneficial. Our patients deserve it. Our conscience demands it,” they wrote.
The full study is available here.
