levels of autonomy

Levels of autonomy are a way of putting a number on how much a robot does for itself versus how much a person does for it. They turn the fuzzy question are you sure it is autonomous? into a clear ladder: at the bottom a human controls every move directly, at the top the robot handles the entire job alone, and in the middle they share the work in various ways. The point is to talk precisely, so that robot A and robot B can be compared instead of both being vaguely called automatic.

A typical ladder runs roughly like this. At the lowest rung the person drives everything in real time, like flying a drone by hand. One step up, the robot helps — it holds a steady hover or keeps a lane while the human still decides where to go. Higher still is shared or supervised control: the robot does most of the work and the person watches, ready to take over or approve key choices. At the top the robot plans and acts entirely on its own, asking nothing of anyone for the whole task.

These scales matter most where safety and law are involved. The best-known example is self-driving cars, often described in six levels from no automation up to full self-driving, with several in-between stages where the human must stay ready to grab the wheel. Naming the level honestly is crucial, because a robot sold as more autonomous than it really is can lull a person into trusting it at exactly the wrong moment.