AI0-001 AI Governance and Ethics Practice Question

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  Demographic parity

  Why it's wrong here

  Demographic parity requires equal positive prediction rates across protected groups, ignoring similarity on the target variable, so it does not match the stated requirement. It suits contexts demanding equal selection rates, whereas consistency-based fairness measures similar outcomes for similar individuals regardless of protected attributes.

• ✗

  Calibration

  Why it's wrong here

  Calibration ensures predicted probabilities match observed outcome rates within groups, not that similar individuals receive similar predictions. The stem describes predictive parity, which compares outcomes conditional on the target variable. Calibration is tempting because it also concerns outcome rates, but it addresses probability reliability rather than parity of predictions across protected attributes.

• ✓

  Individual fairness

  Why this is correct

  Individual fairness requires that similar individuals receive similar predictions, measuring outcome consistency between comparable cases irrespective of protected attributes. This matches the bank's requirement, unlike group fairness metrics such as demographic parity or equalised odds.

• ✗

  Equalised odds

  Why it's wrong here

  Equalised odds equalises true positive and false positive rates across protected groups, matching outcomes conditional on the actual label. The stem instead describes matching predictions for individuals similar on the target variable, which is predictive parity. Equalised odds is tempting because it also conditions on the true outcome, but it constrains error rates, not calibration.