Fairness
Ensuring a model does not produce biased or discriminatory outcomes across groups.
Fairness in AI means a model treats individuals and groups equitably, avoiding outputs that systematically disadvantage protected classes such as gender, race, or age. Training data must be representative, and deployed models should be monitored for performance drift across subgroups over time. Amazon SageMaker Clarify is the primary AWS tool for measuring fairness: it quantifies bias in datasets before training and in predictions after deployment, using metrics like class imbalance and disparate impact. Note that Clarify provides both bias detection and explainability — two distinct capabilities the exam expects you to tell apart.
PlayPrepHQ study notes are written and reviewed against primary exam sources. How we create & review content →