Skip to content

Create a Model Card

A model card is like a nutrition label for your machine learning model. It provides transparency.

What goes in a Model Card?

  1. Model Details: Algorithm type (e.g., Random Forest), date created, developer, version.
  2. Intended Use: What is this built for? (e.g., "Predicting default risk on unsecured personal loans.")
  3. Out of Scope: What should it NOT be used for? (e.g., "Not intended for business loans or mortgages.")
  4. Metrics: Performance across different slices (e.g., Accuracy is 88% overall, but 85% for group A and 90% for group B). Disclosing these differences is crucial for fairness.
  5. Training Data: A brief overview of the dataset. "Trained on 50,000 anonymised loan applications from 2018-2022."
  6. Ethical Considerations: Known limitations or potential biases. "Historically, younger applicants have fewer data points, leading to higher false-positive rates in the 18-21 age bracket."

KSB Mapping

KSB Description How This Addresses It
S5 Deployment, value assessment, and ROI Translating model performance into business impact
S6 Communicate through storytelling and visualisation Presenting ML results to non-technical stakeholders
B4 Consideration of organisational goals Framing technical results in terms of business objectives
B1 Inquisitive approach Exploring creative ways to explain model behaviour