An HR technology company trains a resume-screening model on ten years of its own hiring decisions. A later audit finds the model consistently scores resumes from a particular demographic group lower than equally qualified resumes from other groups, mirroring biased patterns baked into the historical hiring data. Which Responsible AI principle should most directly guide the company's response?

The right principle is fairness, described as ensuring the model treats comparable candidates similarly regardless of demographic group. The audit describes the model producing systematically unequal outcomes for comparable candidates based on group membership, which is precisely what the fairness principle addresses: an artificial intelligence system should not disadvantage any group of people through biased outcomes, and the company must actively test for and correct that imbalance rather than accept whatever pattern the historical data taught the model. The option about increasing server capacity that handles resume uploads is incorrect because server capacity concerns reliability and safety, which covers consistent technical performance under expected conditions, not biased scoring. The option about publishing the model's architecture diagram publicly is incorrect because transparency is about giving people understandable information on how and why the system makes decisions, such as disclosing that artificial intelligence is used and explaining its factors; publishing an architecture diagram does nothing to fix or even reveal fairness problems and is not what transparency requires. The option about assigning one project manager to track feature requests is incorrect because accountability means the people and organizations responsible for an artificial intelligence system are answerable for how it operates and impacts people; assigning someone to track feature requests is project-management work unrelated to answering for biased hiring outcomes. Only fairness targets the specific harm found in the audit.