fairness-audit-preplisted

# Fairness audit prep The playbook's third non-negotiable is that anything scoring or ranking employees or candidates gets a fairness audit. This skill builds the plan. It does not run the audit on real data; that happens in your analytics environment with the right access controls, and the results go through Legal. ## Files this skill needs - `03-governance/ai-use-policy.md` (principle 3, the required elements) - `03-governance/deployer-checklist.md` (monthly monitoring template, disparity ratio row) - `05-notebooks/attrition-risk-modeling.ipynb` (worked example of calibration and disparate impact analysis on synthetic data) - `03-governance/risk-assessment-template.md` (section 4, fairness and bias assessment) - `03-governance/pay-equity-governance.md` (if compensation is anywhere in scope) ## Steps 1. **Define the decision and the outcome variable.** What does the tool output (a score, a rank, a pass/fail, a recommendation), and what real decision does it feed? Name the favorable outcome you'll measure selection rates against (advanced to interview, flagged for retention outreach, recommended for role). 2. **List the groups.** Which protected characteristics are relevant and lawfully available in each jurisdiction in scope? Note where the data isn't collected and what proxy risk that creates. Don't invent group labels the org doesn't have. 3. **Set the pre-deployment tests**, using the checklist in `risk-assessment-template.md` section 4: - Selection rate by grou