Practise vocabulary for HITL, HOTL, automation levels, human oversight, and override communication in AI systems.
0 / 5 completed
1 / 5
Human-in-the-loop (HITL) means:
HITL is the strictest oversight model: the AI suggests, the human approves. Used in high-stakes domains (medical diagnosis approval, legal document review, loan underwriting). The human is in the decision loop, not just monitoring.
2 / 5
Human-on-the-loop (HOTL) is best described as:
HOTL: the AI acts autonomously, but a human monitors and can override. Used in autonomous vehicle monitoring, automated trading oversight, content moderation review queues. Lower friction than HITL but less strict oversight.
3 / 5
The phrase 'if the confidence score falls below 0.7, the decision is escalated to a human agent' describes:
Confidence-based escalation is a common HOTL pattern: the AI handles most cases autonomously (efficiency) but routes uncertain cases to humans (safety). The threshold (0.7) defines the automation/human boundary.
4 / 5
Audit trail requirements for AI-assisted human decisions mean:
Audit trails for HITL/HOTL systems are essential for accountability: logging the AI output, human decision, timestamp, and operator ID allows regulators and quality teams to verify that human oversight is genuinely occurring — not just nominal.
5 / 5
The correct term for the scenario where 'a human reviewer must approve all outputs in the high-risk category before they are sent to the customer' is:
Requiring approval before the decision takes effect = HITL. 'The output is held until a human reviews it' is the defining HITL characteristic — distinguishing it from HOTL where the decision can execute before the human reviews it.