Question

Your team registered an MLflow model that predicts loan approvals. Before deploying it to production, compliance requires you to evaluate whether the model behaves fairly across protected demographic groups such as age and gender. You want to generate this assessment inside your Azure Machine Learning workspace using a supported, low-code approach. Which action best meets this requirement?

Accepted Answer

Create a Responsible AI dashboard component pipeline that includes the fairness assessment on the sensitive features, and review the disaggregated metrics in the workspace UI.

Answer

Enable data drift monitoring on the production endpoint and compare drift metrics between the age and gender feature distributions.

Answer

Log accuracy and F1 score to MLflow during training and compare the aggregate metrics across runs in the jobs UI.

Answer

Deploy the model to a batch endpoint and inspect the scoring script logs for any biased predictions.