Incentive Reflection: When Oversight Reproduces the System It Controls

Oversight fails in two structurally distinct ways. The first is familiar: the overseer is corrupt, captured, or incompetent. The second is less visible and more durable: the oversight layer is placed inside the same incentive field as the system it controls, causing it to reproduce the system's behavior rather than correct it. This note documents the second failure class: Incentive Reflection. The mechanism activates when the overseer's reward structure is aligned with — rather than independent of — the outcomes it is supposed to evaluate. Under this condition, oversight does not weaken over time. It operates at full capacity in the wrong direction. Three cases across healthcare AI and hiring automation instantiate the mechanism. The structural condition that must hold for oversight to function — incentive independence — is derived from the case evidence and formalized as a diagnostic criterion.