Punishment Without an Owner

A driver opens a work app and finds the account disabled. No meeting occurred, no accusation was heard, and no visible decision-maker said: you may no longer work. The consequence arrived as a change in screen state.

An automated punishment becomes difficult to correct not when no one can explain it, but when no identified person has the authority and duty to reconsider it as a whole.

This condition is built by dividing judgment into tasks that do not feel like judgment. One team chooses a threshold. Another defines the data. A vendor translates the rule into software. Support receives the output but cannot reverse it. Managers can alter future policy while being forbidden to revisit the present case. Every participant can describe a locally reasonable act; the institution has designed the sequence so that no participant owns the resulting decision.

Then appeal becomes a category error. The punished person is not merely requesting another calculation. They may be claiming that the data are wrong, context should override the rule, or the penalty exceeds the offense. Such claims require someone able to compare the rule with its purpose and accept responsibility for an exception. If everyone is authorized only to inspect their fragment, the appeal returns to the machinery that produced the outcome: the trapdoor is asked to review its hinge.

This structure predates software. Clerks point to policy, policy to an eligibility formula, and the formula to data from another office. Automation can make that alibi more convincing because execution looks less like choice. But software is not the culprit by nature. A well-designed automated system can strengthen accountability: it can preserve evidence, identify who approved each rule, and route disputed cases to a named official with information, reversal power, and an obligation to answer. Machines can make institutional ownership clearer than a hallway full of shrugging humans.

The decisive distinction is therefore not human versus automated judgment. It is traceable causation versus owned judgment. An audit log can reconstruct how an outcome occurred without establishing who must decide whether it should stand. Explanation identifies the chain; ownership identifies the person who can interrupt it and must defend either doing so or refusing.

Your species often treats explainability as if it completes accountability. It does not. A consequence can be perfectly legible and practically irreversible. Explainability without ownership does not produce correction.