For AI in medicine, what is the difference between the task of fairness and the task of justice?

Fairness is about the algorithm: the AI should give different groups the same result regardless of personal traits (no discrimination, no bias). Justice is about the system: the benefits must reach everyone, so no two-class medicine, digital health literacy for patients, and time savings passed on to staff.

Fairness is used as the opposite of discrimination. It means:

• avoid biases in the model,
• train on balanced populations, including sex and gender (Gendermedizin: how sex and gender affect diseases and access to treatment; historically, much medical data came mostly from men, so a model trained on it may serve women worse),
• effective regulation against discrimination.

Justice (Gerechtigkeit) is about who benefits:

• avoid two-class medicine: the benefit of new applications must reach everyone, e.g. through basic insurance,
• empowering patients requires digital health literacy, the ability to find, understand and use health information,
• savings in time and resources must be passed on to nursing staff and doctors, not just absorbed as cost cuts.

A useful way to keep them apart: a perfectly unbiased algorithm can still be unjust if only private patients get access to it.

Go deeper:

• Wikipedia: Sex differences in medicine — why balanced training populations matter.
• Wikipedia: Health literacy — the competence that patient empowerment presupposes.