Human-in-the-Loop Doesn't Scale - Human-on-the-Loop Does

The article argues that relying on humans to review every action in automated systems (human-in-the-loop) is inefficient and problematic as the volume of actions grows. Instead, it suggests a human-on-the-loop approach, where critical actions are flagged for immediate human review, ensuring potential risks are caught without overwhelming staff. This shift is crucial for maintaining trust and safety in increasingly automated environments, emphasizing context-aware oversight over sheer volume of approvals.