Microsoft's Responsible AI Standard defines fairness as requiring AI systems to treat everyone fairly and avoid affecting similar groups differently — explicitly noting that for use cases like loan applications, the system should make the same recommendations to people with similar financial circumstances regardless of characteristics such as age. A system that automatically rejects every applicant over age 60, regardless of their actual financial qualifications, is discriminating based on a protected attribute (age) rather than evaluating individuals on their merits, which is a direct violation of the fairness principle. This differs from accountability (human oversight and governance of AI decisions), transparency (explainability of how decisions are made), and reliability and safety (consistent, safe technical operation of the system).
Community Discussion