Which metric category helps ensure AI models remain fair and unbiased?

• A. Interpretability
• B. Accuracy
• C. Reliability
• D. Calibration

Answer: (A)

Understanding fairness in AI hinges on being able to see why a model makes the predictions it does. Interpretability provides visibility into the model's reasoning—how features contribute to a decision and how those contributions vary across different groups. With that visibility, you can spot biased patterns, check for disparate impact, and trace decisions back to potentially harmful proxies. This makes it possible to audit and adjust the model to mitigate bias, increasing fairness and accountability. Other aspects like accuracy, reliability, and calibration focus on performance or probability estimates, not on whether decisions treat people fairly. Therefore, interpretability is the category that most directly supports fairness and bias mitigation.