Machine Learning (ML) · Advanced level

Advanced level · Module 2: Interpretability, Fairness and Documentation · Lesson 4 of 4

Model Cards and Risk Management

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

In this lesson you will

  • Write a model card
  • Apply a risk framework to an ML system

Key concepts

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

The lesson

A model card summarises a model: what it does, who built it, training data, intended and out-of-scope uses, performance overall and by group, limits and contact details. It travels with the model.

Risk frameworks such as the NIST AI Risk Management Framework organise work into governing, mapping risks in context, measuring them and managing them. Even small organisations can use their questions as a checklist.

Documentation is part of the product: a model without a card and a named owner should not be deployed.

Step by step

  1. Draft a model card for the credit model.
  2. List risks to applicants and the lender.
  3. Map controls to each risk.
  4. Name the model's owner and review schedule.

Practice activity

Complete the model card and risk assessment for the credit model audit.

Expected result: A model card and risk register a non-specialist manager can understand.

Check your understanding

  1. What does a model card's 'out-of-scope use' section do?

    Show answer

    States uses the model was not designed or validated for.

  2. Name the four functions of the NIST AI RMF.

    Show answer

    Govern, map, measure and manage.

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.