The degree of autonomy designates the extent to which a technical system makes decisions without human involvement. For personalist ontology it is not a purely technical quantity but the decisive ethical threshold: the further the person withdraws from the decision, the more intangible responsibility becomes — up to the responsibility gap.
The four levels
The military and ethical debate (NATO, ICRC, Human Rights Watch) has settled on a four-level taxonomy:
Human-in-the-loop. The system proposes, the person decides. Every single critical action — for instance the target authorization of a weapon or the administration of a medication — requires active human confirmation. The human being is in the decision loop; without their input the system stands still.
Human-on-the-loop. The system acts independently, the person observes and can intervene. The human being is on the loop, that is, above it — monitoring the process and holding a right of veto, but the default motion is autonomous. Critically: the person’s reaction time is limited, and at high event density (drone swarm, stock trading) the veto becomes effectively impossible.
Human-in-command. The person sets the operational framework, rules, and limits but does not intervene in individual actions. The human being is in command — responsible for the whole, not for the single operation. This level is frequently assigned to social interaction robots, care assistance, or driver-assistance systems, where the individual actions are non-critical but the overall scenario must remain under human responsibility.
Fully autonomous. Neither authorization nor monitoring nor framework-setting in the concrete situation. The system itself selects target, means, and execution. In lethal systems this corresponds to the definition of a lethal autonomous weapon system and produces that structural responsibility gap which Sparrow named in 2007.
The personalist-ontological threshold
Between human-on-the-loop and human-in-command there runs a soft boundary; between human-in-command and fully autonomous, a hard one. The reason is not technical but personal: so long as the person sets framework, rules, and goals, they remain present in the event as a moral subject. Once that too falls away, the action is carried out without any personal agency — and this is precisely what contradicts the fundamental insight that only persons can bear responsibility (agere sequitur esse).
The demand for meaningful human control, which the International Committee of the Red Cross and numerous states have brought into the UN GGE LAWS negotiations, aims precisely at this threshold: control must be substantive, not formal — a human being who merely confirms buttons, without being able to understand or examine, does not meet the requirement.
Fields of application
- Military systems: target selection and weapons release — here the threshold is sharpest, because life is at stake (LAWS, combat drone, drone swarm).
- Medical systems: diagnosis, therapy recommendation, medication dosing — human-in-the-loop is the clinical standard here.
- Care robotics: physical assistance mostly human-in-the-loop, social interaction human-in-command (care robot).
- Autonomous driving: SAE levels 0–5 are a field-specific differentiation of the same logic; level 5 corresponds to fully autonomous.
- Financial markets: algorithmic high-frequency trading is effectively fully autonomous; human-on-the-loop survives only as an emergency shutoff.
Ontological classification
Type of: artificial agent, robot
Typically produces in fully autonomous lethal systems: responsibility gap
Chapter assignment: Chapter 5: Oblivion of the Person
Sources: Generated by querying the Personhood ontology.
Further sources:
- International Committee of the Red Cross (2021): ICRC position on autonomous weapon systems. Geneva.
- Human Rights Watch / IHRC (2012 ff.): Losing Humanity — The Case against Killer Robots.
- NATO Science & Technology Organization (2020): Human-Autonomy Teaming — Supporting Dynamically Adjustable Collaboration. Technical Report STO-TR-HFM-247. Neuilly-sur-Seine.
- SAE International (2021): J3016_202104 — Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles.
- Sparrow, Robert (2007): “Killer Robots.” Journal of Applied Philosophy 24(1), pp. 62–77.