Assessment

In this section, we will describe the steps involved in performing a fairness assessment, and introduce some widely (if occasionally incautiously) used fairness metrics, such as demographic parity and equalized odds. We will show how MetricFrame can be used to evaluate the metrics identified during the course of a fairness assessment.

Fairlean provides two primary ways of assessing fairness: MetricFrame, which can be used to perform disaggregated analysis of a particular performance metric (such as accuracy, false positive rate, etc.) across sensitive groups, and a set of predefined fairness metrics that use MetricFrame internally to output an aggregate value. MetricFrame can also be used to output an aggregate value, but the predefined fairness metrics can be used when direct by-group comparison is not necessary. In the Performing a Fairness Assessment, we will dive further into each of these types of fairness assessments.

In the mathematical definitions below, \(X\) denotes a feature vector used for predictions, \(A\) will be a single sensitive feature (such as age or race), and \(Y\) will be the true label. Fairness metrics are phrased in terms of expectations with respect to the distribution over \((X,A,Y)\). Note that \(X\) and \(A\) may or may not share columns, dependent on whether the model is allowed to ‘see’ the sensitive features. When we need to refer to particular values, we will use lower case letters; since we are going to be comparing between groups identified by the sensitive feature, \(\forall a \in A\) will be appearing regularly to indicate that a property holds for all identified groups.

• Performing a Fairness Assessment
  • Identify types of harms
  • Identify the groups that might be harmed
  • Quantify harms
  • Disaggregated metrics
  • Compare quantified harms across the groups
  • Predefined fairness metrics
• Common fairness metrics
  • Demographic parity
  • Equalized odds
  • Equal opportunity
  • The Four Fifths Rule: Often Misapplied
  • Summary
  • References
• Defining custom fairness metrics
• Intersecting Groups
  • Multiple Sensitive Features
  • Control Features
• Confidence Interval Estimation
  • Bootstrapping
  • Bootstrapping MetricFrame
  • Summary
• Advanced Usage of MetricFrame
  • Extra Arguments to Metric functions
  • More Complex Metrics
• Plotting
  • Plotting grouped metrics
• Saving and loading MetricFrame

Fairlearn dashboard

The Fairlearn dashboard was a Jupyter notebook widget for assessing how a model’s predictions impact different groups (e.g., different ethnicities), and also for comparing multiple models along different fairness and performance metrics.

Note

The FairlearnDashboard is no longer being developed as part of Fairlearn. For more information on how to use it refer to microsoft/responsible-ai-toolbox. Fairlearn provides some of the existing functionality through matplotlib-based visualizations. Refer to the Plotting section.