Game setup
Select a mode, then start a round.
Score
0
Question
0/0
Streak
0
Time
—
Canvas matching board
Start a game to receive a fairness concern.
Drag a metric to the target or click a card. Keyboard: 1–6
Ready. Configure the round and press Start game.
Current group evidence
| Group | Total | TP | FP | TN | FN | Selection rate | Accuracy | Precision | Recall | FPR | FNR |
|---|
Interactive group metric calculator
0.50
10 percentage points
Reference value
—
Comparison value
—
Difference / ratio
—
—
Load or generate group data to calculate fairness metrics.
Reference confusion matrix
Comparison confusion matrix
Group performance comparison
Fairness versus accuracy
Threshold performance curves
Disparate impact ratios
Intersectional disparity heatmap
Searchable metric reference library
Trade-off reminder: Fairness definitions can conflict, especially when group base rates differ. Pair metrics with context, uncertainty, stakeholder harms and legal review.
Custom scenario builder
Data quality warnings
- Small group sizes produce unstable estimates.
- Missing demographic labels can bias comparisons.
- Historical labels may encode discrimination.
- Proxy variables can reproduce protected attributes.
- Class imbalance can hide harmful error rates.
Mitigation choices
- Collect better and more representative data.
- Review labels, features and outcome definitions.
- Adjust thresholds with harm-aware evaluation.
- Use fairness constraints where appropriate.
- Add human review, appeals and monitoring.
Performance and fairness audit report
Final score
0
Accuracy
0%
Best streak
0
Average time
0s
Metrics mastered
No completed questions.
Needs practice
No completed questions.
Audit findings
Generate group evidence to create an audit summary.
| # | Industry | Concern | Selected | Correct metric | Result | Seconds |
|---|
Accessibility and display
Progress saves automatically in this browser.