Fairness Audit
Explore an example audit (demo) or audit your own model outputs using a CSV workflow.
How does this tool work? (for non-technical users)
This tool helps you check whether a classification model treats different groups (such as gender or race) differently in its predictions.
What data do I need?
- y_true: the real outcome that actually happened (for example: did someone repay a loan, was income above 50K, etc.).
- y_pred or score: what the model predicted.
- y_pred = predicted class (0 or 1)
- score = probability (between 0 and 1)
- group: the group you want to audit fairness for (for example: Male/Female, race categories, age groups).
Where does this data usually come from?
- y_true comes from historical outcomes in your dataset.
- y_pred / score comes from running your model and exporting its predictions.
- group comes from demographic or categorical fields you already track.
What if I don’t have a model?
You can use the Demo Mode to see how fairness metrics behave, or upload example data using the provided sample CSV template.
🔒 Privacy note: All files are processed locally in your browser. No data is uploaded to any server.
Example Model (Demo)
DemoRun a demo audit on built-in sample data.
Audit Your Model
UploadUpload a main dataset + a predictions file, then merge by an ID column.
Step 1 — Main Dataset CSV
Contains demographics (group) and real outcomes (y_true). Example: a historical dataset.
Step 2 — Predictions CSV
Contains model outputs (y_pred or score) and an ID column to match rows.