AI Ethics and Fairness

Use this skill to audit AI systems for bias, fairness, representation, and governance risk.

Review scope

• The use case and decision context.
• Who may benefit, who may be harmed, and which groups matter.
• Data quality, representation, and proxy variables.
• Model outputs, thresholds, and error distribution across groups.
• Monitoring, documentation, and escalation paths.

Process

1. Define what fairness means for this use case.
2. Identify sensitive attributes and measurement limits.
3. Compare outcomes across relevant groups.
4. Surface tradeoffs between metrics, business goals, and harms.
5. Recommend mitigations, monitoring, and documentation.

Output

• Findings and severity.
• Metrics or comparisons that matter.
• Key uncertainties.
• Concrete next steps.