Beyond the prompt: gender ratio in text-to-image models, with a case study on hospital professions

Key Points

• Identified significant gender ratio bias in text-to-image models, indicating underrepresentation of females.
• The analysis revealed that only 30% of generated images for nursing roles depicted women, showing a clear disparity.
• This observational analysis assessed several text-to-image models specifically focused on healthcare settings.
• Findings suggest a need for improved fairness in algorithmic representations of hospital professions.