📇Write Model Cards for AI Transparency
Stop writing model cards from memory. Walk every section of the Mitchell et al. card with a worked classifier, then critique a real card from a major lab and name what's missing — so transparency becomes a habit, not a deliverable.
Phase 1Why model cards exist and what they solve
Why model cards exist and what they solve
A model card is the nutrition label, not the marketing brochure
7 minA model card is the nutrition label, not the marketing brochure
The Mitchell et al. card has nine sections — and they're not optional
7 minThe Mitchell et al. card has nine sections — and they're not optional
The Gender Shades audit is the canonical 'why we need model cards' story
7 minThe Gender Shades audit is the canonical 'why we need model cards' story
Intended Use only counts if Out-of-Scope is equally specific
7 minIntended Use only counts if Out-of-Scope is equally specific
Phase 2Draft each section for a real classifier
Draft each section for a real classifier
Pick a tiny classifier — sentiment, spam, or species ID — to be the worked example
7 minPick a tiny classifier — sentiment, spam, or species ID — to be the worked example
Section 3 — Factors: name the axes performance could vary on
8 minSection 3 — Factors: name the axes performance could vary on
Section 4 — Metrics: name the measure, the threshold, and the variation method
8 minSection 4 — Metrics: name the measure, the threshold, and the variation method
Sections 5 & 6 — Evaluation Data and Training Data: what was the model fed?
8 minSections 5 & 6 — Evaluation Data and Training Data: what was the model fed?
Section 7 — Quantitative Analyses: unitary AND intersectional, or you haven't done it
9 minSection 7 — Quantitative Analyses: unitary AND intersectional, or you haven't done it
Phase 3Model card vs datasheet vs system card
Model card vs datasheet vs system card
Datasheets describe data, model cards describe models — and you need both
7 minDatasheets describe data, model cards describe models — and you need both
System cards describe deployed products — different audience, different risks
8 minSystem cards describe deployed products — different audience, different risks
Where each document lives in the AI governance stack
8 minWhere each document lives in the AI governance stack
Pick the right document for your audience — and don't conflate them
8 minPick the right document for your audience — and don't conflate them
Phase 4Critique a real card and name what's missing
Critique a real card and name what's missing
Critique a real model card from a major lab and identify what's missing
10 minCritique a real model card from a major lab and identify what's missing
Frequently asked questions
- What is a model card and why do AI teams write them?
- This is covered in the “Write Model Cards for AI Transparency” learning path. Start with daily 5-minute micro-lessons that build from fundamentals to hands-on application.
- What sections does the Mitchell et al. (2019) model card include?
- This is covered in the “Write Model Cards for AI Transparency” learning path. Start with daily 5-minute micro-lessons that build from fundamentals to hands-on application.
- How is a model card different from a datasheet or a system card?
- This is covered in the “Write Model Cards for AI Transparency” learning path. Start with daily 5-minute micro-lessons that build from fundamentals to hands-on application.
- What's the most common mistake teams make when writing a model card?
- This is covered in the “Write Model Cards for AI Transparency” learning path. Start with daily 5-minute micro-lessons that build from fundamentals to hands-on application.
- How do I evaluate whether a model card from a major lab is actually useful?
- This is covered in the “Write Model Cards for AI Transparency” learning path. Start with daily 5-minute micro-lessons that build from fundamentals to hands-on application.
Related paths
🐍Python Decorators Introduction
Build one mental model for Python decorators that covers closures, argument passing, functools.wraps, and stacking — then ship a working caching or logging decorator from scratch in under 30 lines.
🦀Rust Lifetimes Explained
Stop reading `'a` as line noise and start reading it as scope arithmetic — one failing snippet at a time — until you can thread lifetimes through a small parser or iterator adapter without fighting the borrow checker.
☸️Kubernetes Core Concepts
Stop drowning in 30+ resource types. Build the mental model one primitive at a time -- pods, deployments, services, ingress, config -- then deploy a real app with rolling updates and health checks.
📈Big O Intuition
Stop treating Big O as math you memorized for an interview — build the intuition to spot O(n²) disasters, pick the right data structure without thinking, and rewrite a slow function from O(n²) to O(n) in under five minutes.