Machine Learning (ML) · Advanced level

Advanced level · Module 2 of 6 · 22 guided hours

Interpretability, Fairness and Documentation

Explain individual predictions, measure and reduce unfair outcomes and document models so they are used responsibly.

Learning objectives

  • Explain individual predictions with local methods
  • Measure fairness across groups and choose mitigation strategies
  • Document models with model cards and risk assessments

Topics

  • local explanations and SHAP concepts
  • fairness metrics
  • trade-offs between fairness definitions
  • mitigation: data, model and threshold
  • model cards
  • risk frameworks such as the NIST AI RMF

Lessons

  1. Lesson 1 Explaining Individual Predictions

    Explain why a model made a particular prediction for one person or case.

    local explanationShapley valuesfeature contributionsexplanation limits

  2. Lesson 2 Measuring Fairness

    Measure whether a model's outcomes and errors differ across groups, and understand the trade-offs between fairness definitions.

    demographic parityequal opportunityfalse positive and negative rates by groupprotected and proxy attributes

  3. Lesson 3 Mitigating Unfairness

    Reduce unfair outcomes through data, modelling and decision-threshold choices, and check the costs.

    data collection and reweightingconstrained traininggroup-specific thresholdstrade-off evaluation

  4. Lesson 4 Model Cards and Risk Management

    Document models and assess their risks so that others use them appropriately.

    model cardintended and out-of-scope userisk assessmentNIST AI Risk Management Framework

Practical activity

Audit a provided credit-scoring model for a microfinance lender: explain individual decisions, compare outcomes across groups and write a model card with recommendations.

Expected outcomes

  • Explainable decisions
  • Fairness audits and responsible documentation

Assessment

Fairness audit report and model card marked against a rubric.

Taught live by an instructor — online or in person. Free course lessons are for self-study and do not lead to a certificate — see registration options.