The most stable injustice is the one that cannot be attributed.
A civilization can survive many kinds of failure.
It can survive corruption.
It can survive incompetence.
It can survive biased judges and unjust laws.
But there is one failure that changes the type of the world:
when harm occurs, and no one can be named.
Not because the harm was “nobody’s fault” in the moral sense—
but because the structure no longer allows fault to land anywhere.
A mistake still happens.
But the mistake no longer belongs to anyone.
And once that becomes normal, the world becomes strangely invulnerable:
Not to error—
but to accountability.
1) In the old world, injustice still had an owner.
Even in brutal systems, there was usually someone you could point to.
A judge.
A governor.
A priest.
A king.
A committee.
An officer.
A landlord.
A supervisor.
They might be protected.
They might be untouchable.
But they existed as targets of moral language.
You could say:
“He did this.”
“They decided this.”
“This institution is guilty.”
“He did this.”
“They decided this.”
“This institution is guilty.”
That ability matters more than we admit.
It makes outrage intelligible.
It makes resistance coherent.
It makes reform possible.
Because reform begins with a sentence:
“This is unacceptable—and someone is responsible.”
2) The new world doesn’t remove harm. It removes ownership.
Modern systems often boast that they reduce error:
more consistent decisions, fewer biases, better compliance.
But the deeper shift is not about fewer mistakes.
It is about what happens when mistakes still occur.
In complex automated systems, harm can be produced by:
- training data
- model architecture
- threshold settings
- optimization goals
- vendor components
- policy documents
- monitoring dashboards
- human override behavior
- delayed feedback loops
- legal constraints
- organizational incentives
When something goes wrong, the investigation can find “contributors.”
But it struggles to find an author.
Instead of a guilty subject, you get a diagram.
Instead of a responsible person, you get a chain.
So the post-incident conversation becomes strangely weightless: - “There was a failure in the system.”
- “We will improve the model.”
- “We will adjust the threshold.”
- “We will update our processes.”
- “We will retrain staff.”
- “We take this seriously.”
Everything is said.
Nothing lands.
Because none of these sentences contain a name that can bear consequence.
3) Distributed responsibility is not shared responsibility.
People often confuse two things:
distributed responsibility and shared responsibility.
distributed responsibility and shared responsibility.
Shared responsibility means: multiple people each carry moral burden.
Distributed responsibility means: moral burden dissolves because it can’t locate a carrier.
The difference is simple:
In shared responsibility, you can still ask: Who must answer?
In distributed responsibility, the question itself becomes awkward.
Everyone says:
“I only did my part.”
“I followed the process.”
“I used the approved tool.”
“I complied with standards.”
“I didn’t design the policy.”
No one is lying.
And that’s the problem.
The system creates a world where truth becomes a shield against accountability.
4) “The system was wrong” is not the same as “someone was wrong.”
A civilization that cannot attribute fault begins to lose a key human function:
moral repair.
Repair requires an arc:
- error
- confession
- blame
- consequence
- correction
- forgiveness or punishment
But when errors have no owner, there is no confession.
No shame.
No moral consequence.
Only technical iteration.
So wrongdoing becomes an engineering bug.
And victims are told—politely—to accept the update cycle.
This is why the new injustice is so stable:
It is not defended by ideology.
It is stabilized by process.
5) The cruel elegance: the more “objective” the system, the harder it is to contest.
When a human judge makes a terrible call, you can at least attack the person:
bias, incompetence, cruelty, corruption.
When a model makes a terrible call, the critique becomes abstract:
data distributions, statistical error, false positives.
The language of contest becomes technical.
And technical language is not equally available to citizens.
So the system gains a new kind of legitimacy:
Not because it is right,
but because it is difficult to argue against.
Even when you know you were harmed, you meet a wall:
“The model flagged you.”
“The threshold was met.”
“The score was too low.”
“The model flagged you.”
“The threshold was met.”
“The score was too low.”
Those sentences sound neutral.
They quietly deny you what moral argument needs:
an opponent.
an opponent.
6) Once mistakes belong to no one, fear becomes the rational emotion.
In a world where harm has no author, you cannot negotiate with power.
You cannot appeal to mercy.
You cannot shame the decision-maker.
You cannot demand accountability from a person.
So behavior shifts.
People become cautious—not ethically, but statistically.
They avoid anything that may trigger a gate.
They self-censor.
They self-normalize.
They treat deviation as danger.
Not because they have been convinced.
Because they have learned:
if I get hurt, there may be no one to blame—
and therefore no one to repair.
This is how system governance produces obedience without ideology.
Closing: accountability is a civilizational load-bearing beam
A world can handle many injustices.
But when a mistake no longer belongs to anyone, civilization loses something deeper than fairness:
It loses the ability to make moral sense of harm.
And once moral sense collapses, responsibility becomes theater.
You can still have processes.
You can still have audits.
You can still have “improvements.”
But you no longer have a human structure where judgment can be owned, contested, and repaired.
That is why the most dangerous line is not:
“The system makes fewer mistakes.”
“The system makes fewer mistakes.”
It is:
“The system made a mistake—and no one is responsible.”
When that sentence becomes normal, the Empty Chair is no longer a metaphor.
It is the reality of a civilization that still functions—
but can no longer name who must answer for what it does.