A starting point for discussion
The idea
The author applies three categories of human control to systems that use algorithms: the relevant action requires a person’s command; the system can act on its own, but a person can monitor and override it; or it acts without human monitoring or override.
Place in the reasoning
AI assessment · Connected to reasoning
The source distinguishes systems requiring human command, systems subject to monitoring and override, and systems without human monitoring or override; these categories do not establish that oversight is effective.
Computer systems can shape what people do without taking over
The categories clarify that algorithmic systems can involve different degrees of human participation rather than necessarily removing humans.
People need to understand decisions before they can check them
Monitoring and override do not by themselves establish that a human can understand or effectively challenge the system.
What this means and why it matters
What this means
In the first category, a person must command the system to take the relevant action. In the second, the system can act on its own while a person can watch it and stop or change its action. In the third, no person watches or can stop or change what it does.
Why it matters
The categories distinguish between a required human command, later checking, and no human monitoring or ability to stop or change the system. They do not show whether a person can understand or effectively challenge the system.
Conditions & assumptions
Source coverage: Full readable text of the author manuscript from an accessible mirror; not the version formatted by the publisher. The passage adapts three categories from discussion of robotic weapons to other systems, distinguishing them by whether a human command is required and whether a person can watch, stop, or change what the system does. Addresses: The categories of human control
Questions this connects to
AI-suggested relationships. Authors have not necessarily responded to one another.