Treat historical outcomes as potentially biased labels: What decisive requirement should govern interpretation
Historical decisions may encode discriminatory practices; assess their provenance and limitations rather than treating them as objective labels.
The question
An internal employee-assistance model is trained on historical records showing which requests supervisors resolved successfully. Some supervisors routinely denied requests from one group. What decisive requirement should govern interpretation of those labels?
Preparing for AIGP? Take the free 5-min readiness quiz →
- Replace every historical label with a newly collected employee preference.Preferences may provide useful context, but they do not necessarily measure assistance quality or resolve the original outcome bias.
- Treat historical outcomes as potentially biased labels, not neutral ground truth. ✓Historical labels can reproduce earlier discrimination, so their provenance and fitness require examination before they guide model development.
- Use historical outcomes as the target without modification.This treats prior decisions as neutral truth, although inconsistent or discriminatory practices may have shaped the recorded outcomes.
- Retain the labels and rely on aggregate accuracy to reveal any concerns.Aggregate accuracy can conceal subgroup harms and cannot establish that historically generated labels represent fair outcomes.
The trap
Ask who created the label and what decision process produced it before trusting the target variable. How to remember it
Historical decisions may encode discriminatory practices; assess their provenance and limitations rather than treating them as objective labels.
How many of these would you get right?
One of 1581 AIGP questions on Certsqill. Take a free five-minute check and see your score per domain — not one number, but which section to open tonight.
Test your AIGP readiness — freeMore Understanding How to Govern AI Development questions
- Treat the missingness as systematic: Which conclusion is decisive? →
- Evaluate correlated features and subgroup effects: During review, which conclusion is decisive? →
- Use a representative test set untouched by training: What evidence is decisive before deployment? →
- All 426 Understanding How to Govern AI Development questions →
Part of the Certsqill AIGP question bank · Understanding How to Govern AI Development ·
Every answer, right and wrong, comes with its own explanation.
