Detect and Mitigate Model Bias

The Scenario

A recruitment platform uses a machine learning model to score CVs and rank candidates. An internal audit reveals that the model consistently scores male candidates 12% higher than female candidates with identical qualifications. The board wants a full audit and remediation plan.

The Brief

Conduct a bias audit of the hypothetical model. Identify how bias could have entered the system, propose detection methods, and design mitigation strategies that do not simply remove protected attributes.

Deliverables

• A taxonomy of how bias enters ML systems (training data, feature selection, label bias, feedback loops)
• Detection methods: statistical tests and metrics you would use to measure demographic parity and equal opportunity
• Mitigation strategies: at least 3 approaches (pre-processing, in-processing, post-processing) with trade-offs
• An ethical framework for deciding how much accuracy you are willing to sacrifice for fairness

Submit Your Work

Your submission is graded against the rubric on the right. If you pass, you get a public Badge URL you can share on LinkedIn. There is no draft save, so work offline first and paste your finished response here.