← Lessons

quiz vs the machine

Platinum1720

Machine Learning

The Model Cards and Transparency

How structured documentation communicates a model's intended use and limits.

5 min read · advanced · beat Platinum to climb

Documenting a model honestly

A model card is a short structured document that accompanies a released model, describing what it is, how it was evaluated, and where it should not be used. It makes safety and limitations legible to users and auditors.

Typical sections

  • Intended use and out of scope uses.
  • Training data sources at a high level and known gaps.
  • Evaluation results, including disaggregated metrics across groups.
  • Limitations and risks, such as known failure modes and bias findings.
  • Ethical and safety considerations and recommended safeguards.

Why it matters

  • It sets honest expectations so deployers do not misuse the model.
  • It creates accountability and a record for audits and incident response.
  • Disaggregated evaluation in the card exposes uneven performance that an aggregate score would hide.

Pitfalls

  • A card can be vague or omit inconvenient findings, becoming marketing.
  • It is a snapshot, so it must be updated as the model or its use changes.

Key idea

A model card is structured transparency documentation covering intended use, data, disaggregated evaluation, and limitations, creating accountability provided it is honest and kept up to date.

Check yourself

Answer to earn rating on the learn ladder.

1. What is the primary purpose of a model card?

2. Why include disaggregated evaluation in a model card?