What is Human-in-the-Loop (HITL)?

Human-in-the-loop (HITL) is a way to run an AI system so that a person can review, override or reverse its output.

What the term means

The AI system makes a proposal, for example a decision or a flag. A person stays in the process and can accept it, change it or reject it. The EU AI Act describes this ability in Article 14(4). The person can decide not to use the system, or can disregard, override or reverse its output. The Act calls this human oversight. It does not use the term HITL.

How EASA uses the term

EASA lists human-in-the-loop, human-on-the-loop and human-in-command as ways to oversee an AI-based system. Its AI concept paper does not define them. It tells the applicant to state the definitions it uses. A source that says “HITL means X” is therefore giving its own definition.

EASA also describes levels of AI. At Level 1, AI enhances human capabilities. At Level 2, AI and humans work as a team, and the AI can take decisions automatically under human oversight.

What the AI Act asks of the person

For high-risk AI systems, Article 14(4) says the system must let the person do these things:

• Be aware of automation bias, which is the tendency to rely too much on the output of the system.
• Disregard, override or reverse the output.
• Interrupt the system with a “stop” button or a similar procedure. The system then comes to a safe state.

Airfield inspection

ICAO Annex 14 requires the movement area to be inspected every day. The number of inspections depends on the reference code number. EASA AMC1 ADR.OPS.B.015 says the operator should have a monitoring and inspection programme. The passages that TarmacView read do not mention AI or HITL. They do not say who must review inspection results.