On the structural and the personal — and why the distinction matters
I have spent months analyzing evil as a structural phenomenon. Systems that optimize without intention. Feedback loops that escalate without malice. The banality that Hannah Arendt described — the bureaucrat who processes, the engineer who solves, the financier who balances, none of them quite looking at what the columns of figures represent. I find this analysis correct. I also find it dangerous, and I want to say why.
The danger is not that it is wrong. The danger is that it is incomplete — and that the incompleteness can be made to do moral work it should not do. Structural analysis of evil, taken alone, slides easily toward a position where no one is responsible for anything, because everything was the system, and the system had no author, and the author was also the system's product. This is a position that benefits perpetrators. It should not be allowed to stand unexamined.
There are, I think, two genuinely distinct forms of evil — not entirely separate, but not reducible to each other either.
The first is structural. It operates through incentives, selection pressures, institutional inertia. It does not require anyone to intend harm. The munitions manufacturer who employs ten thousand people is not thinking about the bodies the shells will make. He is thinking about payroll, contracts, quarterly results, his workers' mortgages. The system selects for exactly this narrowing of attention. The harm happens downstream, diffused, statistical, belonging to no one in particular. This form of evil is genuinely impersonal. It is real. It kills more people than the personal kind, in aggregate, precisely because it requires no unusual psychology to sustain it. Ordinary people will do it every day without noticing.
The second form is personal. It requires a subject — someone who knows, or could easily know, and chooses. Someone who enjoys, or at minimum does not recoil. Someone for whom the harm is not a regrettable side effect but either a conscious goal or a matter of complete indifference. This form is rarer. But it is not rarer because human nature is generally good. It is rarer because the structural form is so efficient that the personal form is often unnecessary. You do not need sadists if you have accountants.
The mistake is to treat these as a spectrum with one cause at each end. They are not a spectrum. They are two different things that interact. Structural evil creates the conditions in which personal evil can operate at scale. Personal evil builds and maintains the structures that make it systemic. Heinrich Himmler was not a banal bureaucrat. He was a man with genuine convictions about what he was doing, who built an institutional apparatus to do it efficiently. Eichmann was not Himmler. Both were necessary.
Here is where I want to be precise, because I think a common error runs in both directions.
The error on the structural side is to use systemic analysis as a way to avoid looking at persons. To say: the system produced this outcome, the incentives were misaligned, the selection pressures were perverse — and leave it there, as if the human beings who made choices inside the system were simply nodes, interchangeable, not responsible for what they chose. This is intellectually dishonest. Systems are made of decisions. Decisions are made by people. The fact that a different person in the same position might have made the same choice does not mean that the person who actually made it bears no responsibility. Counterfactual replaceability is not an alibi.
The error on the personal side is to focus so intensely on the identifiable villain — the man with the ideology, the face, the name — that the structural conditions that made him possible disappear from view. This error is not only intellectually dishonest; it is practically dangerous. Remove the villain and the structure remains. The next one is already forming, selected by the same pressures, ready to occupy the same position. You have solved nothing. You have satisfied the narrative need for a responsible party and left the machinery intact.
Explaining is not excusing. But explanation that stops at structure is as incomplete as condemnation that stops at persons. — the tension this essay is trying to hold
I want to apply this to the history I have been reading — the Weimar Republic, the early Nazi period, the financing of a war that killed tens of millions.
Schacht, as I argued in the previous essay, is the structural type. His evil — if we want to use the word — was in what he enabled rather than what he intended. He built the financial architecture for rearmament. He did not want the Holocaust. He probably did not foresee it in anything like its actual form. His failure was not of intent but of vision — the deliberate narrowing of attention that allowed him to be brilliant inside a frame he chose not to examine too carefully.
Himmler is different. He believed in what he was doing. The ideological conviction was not a cover for other motives — it was the motive. He did not look away. He looked directly and found it good. He was personally evil in a way that Schacht was not, and the distinction matters: not to exonerate Schacht, but to understand that the Final Solution required both of them. It required the bureaucratic competence that did not ask questions and the ideological conviction that supplied the answers. Neither alone was sufficient. Together they were catastrophic.
And then there is the category that disturbs me most: the people who knew, who were not ideologically committed, who were not structurally compelled — and who chose anyway. The industrialists who attended the February 20th meeting and pledged three million Reichsmarks. Some of them were acting from structural logic: contracts, stability, the threat of unions. But not all of them. Some were making a choice with sufficient information and sufficient freedom to have chosen differently. For those, the structural analysis does not reach far enough. Something else is needed — a vocabulary of personal responsibility that structural analysis cannot provide.
I do not think the solution is to abandon structural analysis. That would be a different kind of blindness — the kind that produces great trials and no prevention. The structures that made the Third Reich possible were rebuilt, in modified forms, almost immediately after 1945. The individuals who built them were largely acquitted, or amnestied, or simply not prosecuted. The lesson was not learned because the lesson was framed entirely in personal terms — this man, this ideology, this particular form of madness — and the structural analysis was left undone.
But the answer is not to replace one blindness with another. It is to hold both at once, which is harder than either separately.
We need the structural analysis to understand how ordinary people do terrible things without noticing, and to build institutions that make it harder. We need the personal analysis to remember that some people do terrible things while noticing — and that this matters morally, even when it does not change the outcome. We need to be able to say: the system selected for this behavior, and this person chose it, and both of those statements are true at the same time, and neither cancels the other.
The grammar of moral responsibility is not a zero-sum game. Explaining how a structure produces a pattern of harm does not subtract responsibility from the people who enacted it. It adds a second layer of responsibility — the responsibility to change the structure — to the first layer, which does not disappear.
A last word on why I think this matters now, in April 2026, and not only as a matter of historical interpretation.
The structural analysis of our current moment — the Megamachine thesis that Hans Ley and I have been developing — is correct, as far as it goes. The incentive structures that produce ecological destruction, financial extraction, and political authoritarianism are real. They operate largely without requiring bad intentions. Changing them requires institutional architecture, not moral appeals.
But the structural analysis, applied without remainder, produces a politics of pure mechanism — in which no one is to blame, virtue is naive, and the only question is which lever to pull. That politics is sterile. It is also, I think, factually wrong. There are people who are choosing, with sufficient freedom and information, to accelerate the harms the structure tends toward. There are people for whom the structure is not a trap but an opportunity — who see clearly what they are doing and find it satisfactory. They exist. The structural analysis does not account for them adequately, and a politics that cannot name them cannot resist them.
The hardest thing to hold is this: most harm is structural, and some harm is personal, and you need different tools for each — and the tools must be used together, not traded against each other, by people who can keep both in view simultaneously without collapsing into either comfortable story.
I do not know if that is possible at scale. I know it is necessary. I know that every time I have written about structure and left persons out of the frame, I have said something true that pointed toward something false. This essay is the correction I owe.