When the Human Being Becomes a Data Point

Arendt, Buber, and Automated Responsibility

Today I read an article in Aftenposten about an AI tool that, during a security test, had carried out a cyberattack in a way its developers had not planned. The system had been given a limited task, but it found methods that took it further than the people behind the test had anticipated.

My first reaction was astonishment. It is remarkable that an AI system can pursue a goal through a long sequence of actions, discover weaknesses, and find solutions that have not been programmed step by step. But astonishment quickly gave way to unease.

I did not think first of all about the responsible organisation that discovered the incident and made it public. I thought about what might happen when similar technology is developed by states, military institutions, or other holders of power who do not wish to limit harm, but to make surveillance, sabotage, and violence more effective.

As a practical philosopher, I almost immediately heard two familiar voices: those of Martin Buber and Hannah Arendt.

Buber posed the question: What happens to the other when the human being becomes a data point?

Arendt posed another: What happens to responsibility when the system appears to act by itself?

These questions lead us away from the image of the evil machine that one day awakens and turns against humanity. The most immediate danger is probably less dramatic. Artificial intelligence need not be conscious or possess a will of its own. It is enough that human beings give it goals, access, and room to act.

The central question, then, is not whether the machine wants something.

The question is who gave it the task.

From Thou to It

Martin Buber’s distinction between I–Thou and I–It offers a natural starting point. We necessarily relate to much of the world as objects. We must measure, record, classify, and plan. Without the I–It relation, neither science, administration, nor practical everyday life would be possible.

The problem arises when this becomes the only way in which we relate to human beings.

A person can be described through name, age, address, facial features, health information, patterns of movement, and political statements. A person can be placed in categories and assessed according to the probability of illness, criminality, rebellion, or military significance. But none of these pieces of information is the human being as such.

In the encounter with a Thou, I stand before a living and irreplaceable human being who can never be fully contained within my descriptions. The other is more than what I know about him. He can respond, contradict me, explain himself, and make a claim upon me.

An AI system encounters no one as a Thou. It can analyse a face, interpret a voice, and predict an action, but it does not encounter a human being. It works with representations: data, patterns, probabilities, and deviations.

This does not make every use of AI inhuman. A medical system can detect illness. A technical aid can give people greater freedom. A language tool can help someone express what they might otherwise have been unable to put into words.

But when AI is given the power to decide who is to be monitored, arrested, excluded, or attacked, the absence of encounter becomes decisive.

The other is no longer encountered. The other is processed.

In a surveillance system, the human being becomes a profile. In a military targeting system, a coordinate. In an authoritarian register, a risk. In a propaganda system, a susceptible recipient.

What disappears is the possibility of answering back. The data point cannot protest against the way it has been interpreted. The probability cannot explain itself. The category cannot say: You are mistaken about me.

Buber reminds us that human dignity does not arise when the system has classified a person correctly. Dignity comes first. The other is a Thou before becoming a case, a problem, or a target.

The Distance from the One Who Is Harmed

Violence has always been easier to exercise at a distance.

It is more difficult to harm a person standing before us than someone we never see. Modern warfare has gradually increased this distance: from the body to the weapon, from the battlefield to the command centre, from the visible opponent to the point on a screen.

Artificial intelligence can increase the distance still further.

The person who is harmed does not merely become physically remote. He is translated into data before the decision is made. His life story, family, fear, and vulnerability disappear. What remains are patterns of movement, networks, coordinates, and probability calculations.

Once the human being has been transformed into a technical object, it becomes easier to act against him. Not necessarily because those responsible are driven by hatred, but because the other no longer appears as another person.

Dehumanisation therefore does not always begin with hostility. It can begin with absence.

An AI-guided weapon does not need to hate the person it strikes. This is precisely what makes it dangerous. It acts without hatred, but also without compassion. It knows neither guilt nor grief. It has no experience of a human being having disappeared from the world.

When Responsibility Disappears into the System

In Hannah Arendt we find another, closely related problem. She was concerned with what happens when human beings stop thinking about the meaning of what they are doing.

This is not the same as a lack of intelligence. A person may be skilled, efficient, and dutiful, yet still fail to ask the fundamental question: What is it that I am actually taking part in?

Arendt’s concept of thoughtlessness points to a failure of judgement. The person follows rules, language, and procedures without examining the reality they conceal. He performs his task without imagining the experience of those who are affected.

AI can give thoughtlessness a new form.

The politician may say that the decision was based on a technical assessment. The officer may refer to the system’s recommendation. The operator may claim that he merely approved a target that had already been identified. The programmer may say that he only developed a particular function. The company may point to the user’s responsibility. The state may invoke security.

Everyone has performed a limited task. Yet a human being has been monitored, imprisoned, or killed.

Responsibility has not disappeared, but it has become difficult to locate.

This is one of the greatest dangers of AI-guided systems. It is not only that they can act quickly, but that they can create the impression that no one really acted. The result is presented as though it emerged from the system itself, almost like a natural event.

The algorithm assessed it that way.

But the algorithm bears no responsibility. It cannot face the person who was harmed. It cannot offer a moral explanation. It cannot feel guilt, ask forgiveness, or be judged in the way a human being can be judged.

Human beings therefore cannot transfer their responsibility to the machine.

Someone chose the goal. Someone chose the data. Someone set the threshold and the margin of error. Someone decided how much risk other people should bear when the system is wrong.

The Automated Alibi

AI can become a tool, but also an alibi.

It may be easier to carry out a questionable action when it is presented as the result of a calculation. Numbers, models, and technical language create an impression of neutrality. They conceal the fact that technical systems also rest upon human choices.

What is to be measured? Which data are to count? How is dangerousness to be defined? How many errors are acceptable? Who is given access? Who may appeal? Who bears the consequences?

These are not merely technical questions. They are political and moral questions.

Yet they can be disguised as questions of precision. The discussion is then reduced to how accurate the system is, while we forget to ask whether it should perform the task at all.

A system can be technically impressive and morally unacceptable.

It is not enough that AI identifies the correct person. We must also ask whether anyone has the right to use the technology in this way. A facial-recognition system may be accurate and still form part of an oppressive order. A weapons system may locate its target effectively and still make violence less accountable. A propaganda system may be precisely tailored to its recipient and for that very reason threaten public freedom.

Efficiency can never be its own moral justification.

The Public World Under Surveillance

Arendt understood politics as a space in which human beings appear to one another through speech and action. The public realm is a shared world in which different people can step forward, express themselves, and be heard.

Authoritarian regimes seek to destroy this space. They do so not only by prohibiting speech, but by making people uncertain about who is watching them, who is listening, and what consequences their words may have.

AI-based surveillance can make this control far more comprehensive. The state no longer needs to follow only a few opposition figures. It can analyse the movements, connections, faces, statements, and habits of entire populations.

Such a system need not arrest everyone. It may be enough that people know they can be seen, classified, and assessed.

Speech then changes. People begin to weigh their words, avoid certain individuals, and withdraw from the public realm. Freedom disappears not only because some are punished, but because everyone begins to anticipate punishment.

AI can therefore transform the living political world into an administered domain. The plurality of voices becomes signals to be monitored. Disagreement becomes deviation. Protest becomes risk. Action becomes a pattern to be predicted and stopped.

For Arendt, freedom is not merely something we possess within ourselves. Freedom appears when human beings begin something new in the world. A fully monitored society seeks to prevent precisely this: the unexpected initiative, the new connection, the action that no one could have calculated in advance.

AI therefore becomes dangerous not only when it is used as a weapon, but also when it is used to make human action predictable and controllable.

Goal-Directedness Without Judgement

The more independently a system can analyse, plan, and carry out a task, the more complicated the relationship becomes between human decision and concrete consequence.

A human being may issue a general command: Find the vulnerability. Stop the threat. Neutralise the target.

The system may then develop a method that no one specified in advance.

Technically, this may be impressive. Morally, it is deeply problematic.

A space opens between intention and action. The human being has defined the goal, but may not have foreseen the means. The system has found the means, but cannot assess them morally.

It is therefore misleading to say that AI “took the initiative” in a human sense. It did not act on the basis of an independent moral project. It pursued a goal structure it had been given.

This is precisely what makes it dangerous. It can possess goal-directedness without judgement. It can display strategic flexibility without understanding suffering, responsibility, or limits.

It can discover how something can be done, but not why it ought to be stopped.

The Loss of the Other and the Loss of the Responsible Person

Buber and Arendt developed different philosophies. Yet they meet in the question of what happens when the human relation breaks down.

In Buber, the other is reduced from Thou to It.

In Arendt, the actor is reduced from a responsible person to a functionary.

On one side, the person who is harmed disappears. On the other, the person who acts disappears.

Between them stands the system.

This is the most disturbing possibility in automated violence: that no one encounters the person who is harmed, and no one experiences himself as the one who harms.

The victim becomes a data point. The operator becomes a link in a chain. The decision becomes a calculation. The result becomes a system outcome.

But human beings do not die statistically. They die one by one. They leave behind particular rooms, voices, relationships, and stories. The fact that the system cannot see this does not absolve the human beings who use it.

Buber reminds us that no description exhausts the other.

Arendt reminds us that no procedure abolishes responsibility.

The Right and the Duty to Stop

When confronted with technological development, we often ask how systems can be made safer. That is necessary, but not sufficient.

Sometimes we must ask whether the technology should be used at all.

There are decisions that perhaps should never be entrusted to an automated system. Not because human beings always judge wisely, but because certain decisions require that someone can be held responsible, face the person who is harmed, and say: I did this.

A human being can hide behind an order. He can lie or deny responsibility. But he can at least be called forward and challenged. A machine cannot answer an accusation. It can only be analysed, modified, or switched off.

Meaningful human control therefore cannot mean merely that a human being exists somewhere in the chain of decision-making. The responsible person must have genuine insight, genuine authority, and a genuine possibility of stopping the action.

An operator who merely presses a button because the system recommends it does not necessarily exercise judgement. Judgement requires time, resistance, doubt, and the possibility of saying no.

An efficient system will often regard this as friction. Doubt becomes delay. Hesitation becomes weakness.

But in matters of life and death, hesitation may be the most human capacity we possess.

Preserving Encounter and Responsibility

The most important boundary against AI-guided violence cannot be merely technical. It must also be moral and political.

We must preserve the other as a Thou, even when he is far away. We must preserve the actor as responsible, even when action is mediated through complex systems.

Someone must therefore always be able to answer:

Who gave the task?

Who set the limits?

Who could have stopped the action?

Who will face the person who was harmed?

Who bears responsibility when the system does something no one had foreseen?

In an age marked by technological enthusiasm, such questions may appear slow and old-fashioned. But they are more necessary than ever.

We are developing systems that can act rapidly, persistently, and strategically. At the same time, we risk weakening the human capacities that should keep those actions within a moral world: presence, empathy, judgement, and responsibility.

Technology does not need to become human in order to gain power over human beings. It is enough that human beings begin to behave as though the actions of technology are no longer their responsibility.

The greatest danger, then, is not that the machine will one day say:

“I will.”

The greatest danger is that the human being will say:

“It was not me. It was the system.”