Governance

modes of human involvement (human-in/on/over-the-loop)

Human-in/on/over-the-loop models describing levels of human intervention.

Definition

A set of oversight models describing degrees of human engagement with AI decision-making: human-in-the-loop (human directly makes or approves decisions), human-on-the-loop (human monitors and can intervene in automated processes), and human-over-the-loop (human provides high-level governance, policy and override capability). Selection is risk-informed based on severity and reversibility of potential harm and operational feasibility.