Portrait of a Surgeon: Artificial Intelligence Reflections.

Area of Science:

• Medical AI
• Computer Vision
• Societal Impact of Technology

Background:

• Text-to-image artificial intelligence (AI) models are increasingly popular tools for generating visuals from textual prompts.
• AI models trained on internet data may inherit and perpetuate societal biases, similar to large language models.
• Investigating AI's potential to reflect or reinforce stereotypes in healthcare professions is crucial.

Purpose of the Study:

• To examine if three common text-to-image AI systems reproduce stereotypes associated with surgeons and healthcare professionals.
• To assess the diversity (race and gender) in AI-generated images of healthcare professionals.
• To understand the implications of AI in healthcare and its potential to reinforce historical biases.

Main Methods:

• Querying three distinct text-to-image AI platforms with prompts related to surgeons and healthcare professionals.
• Analyzing the generated images for common professional attributes (attire, equipment, settings).
• Evaluating the visible race and gender diversity within the AI-generated professional depictions.

Main Results:

• All queried AI platforms successfully depicted common professional elements for surgeons and healthcare workers.
• Significant differences were observed between AI systems regarding visible race and gender diversity in generated images.
• The AI systems tended to underrepresent diversity, potentially reinforcing historical stereotypes.

Conclusions:

• Text-to-image AI systems can perpetuate historical stereotypes of surgeons and healthcare professionals.
• The lack of diversity in AI outputs highlights the need for bias mitigation strategies.
• Understanding and addressing AI biases is critical as these technologies become more integrated into healthcare contexts.