Human-in-the-loop

Content sourced from Wikipedia, licensed under CC BY-SA 3.0.

Human-in-the-loop (HITL) means a system still needs a person to act or decide. It appears in modeling and simulation (live, virtual, and constructive) and is also discussed in the context of weapons that can operate without a human. In machine learning, HITL means people help the computer pick the most important data so the model learns better than by random sampling.

In simulations, a human is part of the scenario and directly influences the outcome, making it hard to reproduce exactly. This helps identify problems or requirements that automation alone might miss. HITL is often called an interactive simulation—for example, flight or driving simulators where a person controls the vehicle. The goal is to train people and to see how changing a process would affect real events.

A real-world use is the U.S. Federal Aviation Administration, which uses HITL to let air traffic controllers test new automation with simulated traffic while watching how the new procedures work. Because a human is involved, HITL can reveal human errors and limits that fully automated testing might miss.

While automation can handle many tasks, humans are still needed to interpret information and make final judgments based on experience. Pure computer simulations are predictable and repeatable but may miss important human factors.

Early planning can benefit from tabletop HITL exercises, but the big design choices usually require a human-in-the-loop simulation. In virtual simulations, HITL exercises test motor skills, decision making, and teamwork.

Three ways people control autonomous weapons were described in a 2012 Human Rights Watch report by Bonnie Docherty.