37. A model recommends which service locations receive extra staff. How would you evaluate it before it affects allocation?
HardMachine learning basics
Technical
Show model answer
Clarify the objective, constraints and consequences for underserved locations. Historical outcomes reflect previous staffing decisions, so a model trained on them may reproduce underinvestment. Check data coverage, leakage and performance by relevant operational groups.
Compare the recommendation with current practice and simpler rules using realistic capacity constraints. Involve service owners, test a monitored pilot with human review and define appeal or override paths. Evaluate service outcomes and distributional effects, not prediction accuracy alone.
What interviewers listen for
- Historical policy bias
- Constrained policy evaluation
- Human review and overrides
- Distributional outcomes