The decision you didn't know you made
Automation acts in your name through defaults and standing rules — producing consequential decisions you are accountable for but never consciously made.
There is a strange new category of liability taking shape, and most of us are already inside it without having noticed the door. It is the decision attributed to you that you did not knowingly make. Not one you made carelessly, or under pressure, or with bad information — those are old and familiar, and the law has words for each. This is something else: an outcome that the world will lay at your feet, that bears your name and your authority, and that you never sat down and chose. It was produced by a default you accepted years ago, a standing rule you set once and forgot, or a system you delegated to and stopped watching. The decision is yours. The making of it was not.
We have built our entire moral vocabulary around the moment of choosing. We praise and blame, credit and convict, on the assumption that behind every consequential act there was a person who weighed it and committed to it. Automation quietly dissolves that assumption while leaving the vocabulary intact. The act still issues from your account, your policy, your configured agent. But the weighing happened long ago, or in the abstract, or never — and the gap between the moment you are answerable for and the moment you were actually present has become wide enough to fall into.
How a choice gets made in your name
The mechanisms are mundane, which is exactly why they are dangerous. A default is a decision someone else made that becomes yours the instant you do not override it; you are bound by it not because you chose it but because you failed to notice it was choosing for you. A standing rule is a decision you made once, in conditions you understood, that keeps firing in conditions you no longer do — the rule does not age, but the world it was written for does. And a delegated system is a decision-maker you authorized to act without checking back, which means every output it produces arrives pre-attributed to you, fully formed, while you were elsewhere. In each case the structure is the same: your authority is present at the moment of action, and your attention is not.
What makes this the strangest kind of liability is that the usual defenses do not reach it. You cannot say you were careless, because there was no moment of care to skip. You cannot say you were misinformed, because you were not informed at all — you were absent. You cannot even say you would have chosen differently, because choosing was never on offer at the moment that mattered. The decision slid out of the system in your name while you were, in every meaningful sense, not there. And "I was not there" is, perversely, the one thing that does not excuse you, because being there was your job and you delegated it away.
The defaults you never read are still your decisions. The world will not ask whether you chose them. It will only ask whether they were yours.
The only protection is a record of what became yours, and when
If you cannot prevent acting through systems — and at scale, you cannot; that is what scale means — then the protection is not to pretend you are present for every act. It is to make the moments of actual authorship legible. Somewhere there was a real decision: someone set this default, adopted this rule, granted this delegation, and at that moment a human choice happened that the later automated acts merely execute. That moment is where accountability genuinely lives, and it is precisely the moment that almost never gets recorded. We log the output and lose the authorization. We capture the act and forget the act that made the act possible.
So the record that matters is not a transcript of every consequence. It is an account of what was set, by whom, and at what point it became your decision rather than the vendor's, the predecessor's, or no one's. When a default is accepted, that acceptance is a decision and deserves a record of its own — who could have changed it, what they were shown, and the fact that they let it stand. When a standing rule keeps firing, the record should carry the moment it was authored and the conditions it was authored for, so that a rule outliving its world can be caught before it does its damage rather than excused after. When a system acts under delegation, the chain back to the human who granted that delegation is the only thing standing between accountable authority and a decision that belongs to nobody.
The alternative is the situation we are drifting toward by default: a world thick with consequential decisions, each one firmly attributed to a person, none of them traceable to a moment that person was actually present for. That is not accountability. It is its costume. It lets institutions enjoy the efficiency of acting through automation while keeping a human name on the line to absorb the blame, and it lets the human discover, only at the moment of harm, that they have been signing things in their sleep. The way out is not to stop delegating. It is to insist that every delegation leave a mark — so that when the question finally comes, who decided this, there is a true answer, and it points to a moment a person could have stood behind, rather than a system no one was watching and a name no one remembers attaching.
— Dispatches · Summit Cognitive
Continue from here
Turn the argument into a practice.
Get new dispatches, assess how your organization handles consequential decisions, or explore Summit Cognitive.