The field of medical AI is shifting towards a more nuanced understanding of the complex interplay between social inequality and medical imaging. Recent studies have demonstrated that deep learning models can detect subtle traces of socioeconomic status and other social factors from medical images, challenging the assumption that these images are neutral biological data. This new direction in research has significant implications for fairness and accuracy in medical AI, as it highlights the need to interrogate and disentangle the social fingerprints embedded in clinical data. Noteworthy papers in this area include:

• A study showing that algorithms trained on normal chest X-rays can predict health insurance types, a strong proxy for socioeconomic status, with significant accuracy.
• A paper proposing a debiasing framework for computerized adaptive testing, which substantially improves both the generalization ability and fairness of question selection.