Equal Opportunity
You'll learn how equal opportunity fairness ensures all protected groups have equal chances of receiving positive outcomes when they deserve them.
Core Idea
Equal opportunity is a fairness criterion that focuses specifically on true positive rates (TPR) across different protected groups. Unlike demographic parity (which requires equal overall positive prediction rates regardless of ground truth), equal opportunity requires that among individuals who actually belong to the positive class, the model should predict positive outcomes at equal rates across all groups.
In plain terms: if two qualified candidates from different demographic groups both deserve a loan, job, or medical treatment, the model should approve them at the same rate.
The Mathematical Requirement
For protected groups A and B:
P(Ŷ = 1 | Y = 1, Group = A) = P(Ŷ = 1 | Y = 1, Group = B)
This means: given that the true label is positive, the probability of a positive prediction should be the same regardless of group membership.
Why It Matters
Equal opportunity is particularly important when the cost of false negatives is high. Consider:
- Medical diagnosis: Missing a disease (false negative) in one demographic group more often than another is ethically problematic
- Loan approval: Denying credit to qualified applicants from certain groups perpetuates inequality
- College admissions: Rejecting qualified students from underrepresented groups limits opportunity
This criterion focuses fairness efforts on those who should receive the positive outcome, rather than equalizing predictions across everyone (including unqualified individuals).
Comparison to Demographic Parity
While demographic parity demands equal positive rates overall, equal opportunity only constrains the TPR. You might still have different false positive rates across groups—the model could be more cautious with one group's negative cases without violating equal opportunity.
Key Takeaway: Equal opportunity fairness requires that qualified individuals from all protected groups have equal chances of being correctly identified as positive, focusing fairness on deserving recipients rather than the entire population.