User Guide

• 1. Fairness in Machine Learning
  • 1.1. Fairness of AI systems
  • 1.2. Types of harms
  • 1.3. Fairness assessment and unfairness mitigation
    • 1.3.1. Group fairness, sensitive features
    • 1.3.2. Parity constraints
    • 1.3.3. Disparity metrics, group metrics
• 2. Assessment
  • 2.1. Metrics
    • 2.1.1. Ungrouped Metrics
    • 2.1.2. Metrics with Grouping
    • 2.1.3. Scalar Results from MetricFrame
    • 2.1.4. Control features for grouped metrics
  • 2.2. Fairlearn dashboard
    • 2.2.1. Setup and a single-model assessment
    • 2.2.2. Comparing multiple models
• 3. Mitigation
  • 3.1. Preprocessing
  • 3.2. Postprocessing
  • 3.3. Reductions
    • 3.3.1. Fairness constraints for binary classification
    • 3.3.2. Fairness constraints for multi-class classification
    • 3.3.3. Fairness constraints for regression
    • 3.3.4. Exponentiated Gradient
    • 3.3.5. Grid Search
• 4. Migrating from prior versions
  • 4.1. Migrating to v0.5.0 from v0.4.6
    • 4.1.1. Metrics
    • 4.1.2. Renamed object attributes
    • 4.1.3. Exponentiated Gradient and Moments
• 5. Further Resources
  • 5.1. Books
  • 5.2. Papers
  • 5.3. Online demos, talks, tutorials