The Ceremonial Human: Responsibility Without Control
Synthocracy Institute · Method / Working Paper · Research strand · by Martin Novak · July 2026
Definition. The ceremonial human is a person who formally approves, signs, and bears responsibility for a decision that an AI system has, in practice, already made. They hold the accountability. They do not hold the control.
Key points
- The ceremonial human is the failure mode of “human oversight”: a person kept in the process to sign, not to decide.
- It is not a new observation. It builds on Madeleine Clare Elish’s moral crumple zone, the literature on automation bias, and the debate over meaningful human control.
- The Institute’s contribution is a plain public term, a diagnostic test, and a sector-by-sector map — so the pattern can be named, checked, and challenged.
- The danger is a distinctive one: not that a human is absent, but that a human is present to absorb blame for a decision they could not change.
- Oversight is only real when the human has a meaningful ability to do otherwise — to see, to reach the facts, and to overrule in time.
What is the ceremonial human?
Whenever an AI system touches a consequential decision, we are reassured that “a human is in the loop.” Often that is true in the way that matters. Often it is theatre.
The ceremonial human is the theatrical version. The options were narrowed, ranked, scored, and framed by a system before they reached the person. The person’s remaining job is to confirm. They sign; they carry the legal responsibility; and if the outcome is wrong or unfair, they are the one held to account — for a decision they never really made.
We did not take the human out of the loop. We took the decision out of the human.
This is the sharp edge of synthocracy. Power does not leave the room; it moves upstream of the person who appears to hold it. The signature stays where a decision used to be.
This is not a new observation — and that matters
Honesty is part of the method, so we name the lineage. The ceremonial human is a public-facing name for a pattern serious scholars have described for years:
- The moral crumple zone. Madeleine Clare Elish (2019) showed that in complex automated systems, responsibility is often mis-attributed to the nearest human operator — who absorbs the moral and legal force of a failure the way a car’s crumple zone absorbs a crash. The zone protects the system at the expense of the person.
- Automation bias. A long line of human-computer-interaction research shows that people over-trust automated recommendations and “fall asleep at the wheel” — and that prolonged reliance erodes their ability to catch errors at all. Clinical studies have found practitioners follow flawed AI advice even against their own judgment.
- Meaningful human control. The autonomous-systems literature stresses that real control requires an interplay of ability, authority, information, and accountability — and critics note that “human oversight” as written into policy suffers from an inherent imprecision: there is no agreed line between meaningful oversight and rubber-stamping.
Our contribution is not to discover the pattern. It is to give it a term the public can use, a test anyone can apply, and a map of where it hides.
Where the ceremonial human appears
The role recurs across sectors, each with the same shape — a human signing after the system has decided:
- The official who approves a benefit, a tax flag, or a migration case after an algorithmic risk score has sorted it.
- The recruiter who “chooses” from a shortlist, having only ever seen the candidates the system let through.
- The clinician who signs off a care pathway an AI tool has already assembled.
- The judge or court officer handed a risk summary and a recommended disposition.
- The manager who “decides” from a dashboard whose metrics, thresholds, and framing were set elsewhere.
- The moderator who confirms, at volume, the content decisions a model has already made.
In each case the person is real, well-intentioned, and accountable. And in each case the decision arrived pre-compressed.
Why it happens
Three forces produce the ceremonial human, and none require bad faith.
Automation bias. When a system is usually right and always fast, deferring to it becomes the rational default — until deferring is all that happens.
Option compression. By the time a human sees a decision, the system has already discarded most of the possibility space. You cannot meaningfully choose among options you were never shown.
Incentives. For the institution, a human at the end of the process is useful in a specific way: it allows the system to be promoted as superhuman while a person is positioned to absorb the blame if it fails. Oversight becomes, in effect, a liability shield — a way to have it both ways.
The test: genuine oversight or ceremony?
The presence of a human proves nothing. The question is whether that human has a meaningful ability to do otherwise. Four checks:
- Can they see why? Is the system’s output explained in terms the person can actually evaluate — not just a score?
- Can they reach the facts? Do they have access to the underlying case, not only the summary the system produced?
- Do they have the authority, time, and information to overrule? Or is override theoretically available but practically impossible within the workload?
- Is overruling a real, used option? If no one ever departs from the system’s recommendation, “oversight” is a formality, whatever the org chart says.
Where the answers are no, the loop is decorative — and the human in it is ceremonial.
The accountability trap
The ceremonial human creates a specific injustice: responsibility without control. The person carries the liability precisely because they lack the power to have prevented the harm.
The 2018 Uber self-driving fatality in Tempe, Arizona is the textbook case: after the vehicle killed a pedestrian, the safety backup driver was charged and ultimately pled guilty, while the systemic and design factors receded from the story. The nearest human became the crumple zone.
Law has tried to require human oversight — the EU AI Act mandates it for high-risk systems; the GDPR gives a right to human intervention in significant automated decisions. But these requirements can be satisfied on paper while failing in practice: a human is assigned, a box is ticked, and the oversight is ceremonial. A right to “human review” is empty if the reviewer only re-reads the machine’s answer.
What real oversight requires
The fix is not more humans in more loops. It is placing human command where it can still change the outcome:
- Human-in-command, not human-in-the-loop — a person with authority before the options are compressed, not a signatory after.
- Cognitive forcing functions — designs that require active human engagement rather than passive confirmation, to counter automation bias.
- A record and a real off switch — so oversight can be reviewed, and so a human can actually stop or reverse the decision.
This connects the ceremonial human to two other Institute methods. A person who cannot stop or reverse a decision is paired with the problem of the AI kill switch. And the deeper remedy is admissibility: if a system’s decisions cannot be meaningfully overseen by a human with standing, that is a reason to question whether it should have been admitted to act at all.
→ For the concept this sits within, see What Is Synthocracy?. To test any oversight claim, use The Ten Questions.
FAQ
What is a ceremonial human? A person who formally approves and bears responsibility for a decision an AI system has effectively already made — holding the accountability without the control.
How is the ceremonial human different from “human in the loop”? “Human in the loop” describes a design where a person is present. The ceremonial human describes what that person becomes when their role shrinks to confirmation: present, accountable, but unable to change the outcome.
Is this the same as Elish’s “moral crumple zone”? It is closely related. The moral crumple zone describes how the nearest human absorbs blame for a system they couldn’t control. “Ceremonial human” is a plain-language term for that person’s role in everyday AI-mediated decisions, paired with a test for spotting it.
How do I tell real oversight from ceremonial oversight? Ask whether the human can see why the system decided, reach the underlying facts, and overrule it with real authority and time — and whether anyone ever actually does overrule it.
Why does the ceremonial human matter for accountability? Because it produces responsibility without control: the person is liable precisely where they had no power to prevent the outcome, while the system and its designers escape scrutiny.
Martin Novak is the founder of the Synthocracy Institute, an independent research institute studying how decision-making power shifts through AI systems. Warsaw, operating internationally. This method builds on the work of Madeleine Clare Elish and the wider literature on automation bias, meaningful human control, and the responsibility gap
