model cards

Structured documentation for ML models covering intended use, performance metrics across demographics, training data, limitations, and ethical considerations.

Syntax

ai-ethics

# Model Card sections:
# Intended Use | Training Data | Evaluation | Limitations | Ethical Considerations

Example

ai-ethics

# Model Card Template:
"""
## Model Card: Loan Risk Classifier v2.1

### Intended Use
- Assist (not replace) human loan officers
- Not for use in jurisdictions requiring explainability by law

### Training Data
- 500k loan applications (2015–2023)
- Excludes: race, religion, national origin

### Performance
| Group | Accuracy | False Positive Rate |
|-------|----------|--------------------|
| All   | 87%      | 8% |
| Low-income | 83% | 12% | ← disparity noted

### Limitations
- Performance degrades on applicants from states with <1000 training samples
"""