Examining three popular AI text-to-image generators, the study assessed the demographic accuracy of surgeon representations. It found these models overwhelmingly portrayed surgeons as white and male, reinforcing societal biases. Trainees, especially non-white and female, were underrepresented. While one model reflected attending surgeons accurately, all underestimated trainee diversity. The study emphasizes the need for safeguards and feedback mechanisms to prevent AI generators from perpetuating stereotypes, particularly in fields like surgery.
Journal Article by Ali R, Tang OY (…) Asaad WF et 13 al. in JAMA Surg