Parks adds human review before publishing AI-generated trail-closure posters. Which comparison best shows the control’s effect on AI risk?

Human review before publishing AI-generated trail-closure posters is a risk-reducing control. Its effect is shown by comparing inherent risk before the control with residual risk after human review: lower likelihood or impact of incorrect public closures demonstrates that the control works. That before/after comparison is the standard way to justify and communicate treatment effectiveness for parks AI content workflows. Comparing residual risk after human review with inherent risk after human review to show consistency does not isolate the control's effect because inherent risk is the before-control baseline; once human review is applied, the after measure is residual risk, so two after-review ratings cannot demonstrate reduction from the control. Comparing inherent risk before the control with inherent risk after human review to show reduction is incorrect because a control does not produce a new inherent risk; the after-review measure is residual risk, so an inherent-to-inherent comparison cannot demonstrate that human review lowered risk. Comparing residual risk before the control with residual risk after human review to show reduction mislabels the baseline because before the control there is only inherent risk; without an inherent-to-residual comparison, it cannot demonstrate that human review reduced publication risk. Exam caveat: residual risk can remain above appetite even after human review, which would require additional treatment-not a claim that the comparison method is wrong. Operational check: in the risk file for trail-closure generation, record inherent rating, control description (human review), residual rating, and the delta that justifies continued use of the control.