责任的稀释：当没有人需要负责时，坏事是怎么被做出来的
Division of labor split the work, and it split the blame — that second cut was never properly accounted for. What needs repairing is not a return to a pre-division world but a chain of responsibility that survives the split
A wrong that needs no demon
In January 1942, senior Nazi officials met at a villa on the edge of Berlin to discuss what they called the final solution of the Jewish question. The minutes read like a standard administrative document: the participants were deputy ministers and department heads, and what they negotiated was the division of jurisdictions, the handling of legal procedure, the allocation of transport quotas. Nowhere in the record is there a single word objecting to the purpose the meeting existed to serve. After the war, many of those present offered the same defence in court — I was only doing my duty, following the rules.
This is not an isolated case. The gravest atrocities of the twentieth century were rarely carried out by fanatics alone; far more often they were manufactured by large numbers of ordinary administrators, engineers, accountants, and civil servants. They went to work on time, finished their assigned tasks, and went home to their children. The problem then becomes one that moral intuition cannot settle: if everyone performed only their own small segment of the work, where exactly does responsibility land?
The question is hard because it offends two intuitions at once. One says that someone must be answerable. The other says that in fact no one did the whole thing. Both are correct, and the gap between them is what this essay is about.
Where responsibility originally lived
Before organized harm existed, responsibility and the person who acted were the same thing. In Book III of the Nicomachean Ethics, Aristotle supplied a criterion still in use today: only an act that originates in the agent and is performed with knowledge of the circumstances deserves praise or blame. The implicit structure of that criterion is that action, knowledge, and intention must all reside in one and the same person for responsibility to hold. This is not a formal rule; it is a direct answer to the question "who did it?"
Hobbes proposes the opposite pole in Leviathan. Since individuals cannot secure their own safety alone, they reduce the will of all to a single will and hand it to one "person" to represent them, who then bears the whole of judgement and responsibility. This is a project of centralizing responsibility: it does not ask anyone to evade responsibility but gathers it into one identifiable point, so that the question always has an addressee.
Modern organizations do neither. They are not Aristotelian — the three elements are not united — and they are not Hobbesian — responsibility is not gathered into one point. Instead the three are separated and distributed: those who decide know they are deciding but not the details of execution; those who execute know the details but believe they are merely executing; those who record, authorize, and review each hold one segment. Responsibility is neither centralized nor abolished; it is diluted. This is not the moral failure of any individual. It is a structural effect.
The second cut: responsibility broken into operations
The pin factory in Adam Smith's account divided operations. The modern organization divides a second thing: responsibility itself. Once that cut is made, responsibility disappears along at least three paths, and they reinforce one another.
First, each segment carries only local rationality. What each role is asked to do is correct within its own boundary: finance wants costs controlled, legal wants procedure observed, operations wants delivery on schedule. The sum of local optima is not global sanity — which is precisely why the most absurd outcomes are often produced jointly by the most competent people. Philip Kuhn's account of the 1768 sorcery panic in Soulstealers is a specimen of this mechanism: the panic was amplified level by level along the bureaucratic system's routinized reporting and interrogation procedures, every official acting by the book, and no single link in the chain can be named as the culprit. Kuhn's focus is not on a villain but on the amplifying effect of routinized procedure itself — it relieves every participant of responsibility for the whole, because the whole was never anyone's assignment.
Second, information is segmented, so "I did not know" becomes a truthful defence. When each person holds only their own segment, they are not pretending to be ignorant of the whole; they genuinely are. This makes attribution extraordinarily difficult: the law requires proving that someone "ought to have known," and the design of the division of labour is exactly what makes "ought to have known" fail to hold. Stanley Milgram's obedience experiments, published in 1963, offer indirect evidence: the participants who administered shocks were not lacking in moral feeling; what they lacked was the information and the distance needed to reassemble act and consequence — the learner was in another room, and the experimenter took responsibility. Once a structure moves consequences out of view, "I did not know" ceases to be an excuse and becomes an accurate description.
Third, replaceability dilutes the weight of the individual. If someone else in this seat would have done the same thing, then this person's choice appears to matter very little. This is the core of the judgement Hannah Arendt offered in Eichmann in Jerusalem and which has since been widely misread: Eichmann's horror lay not in his being exceptionally wicked but in his being exceptionally thoughtless — he substituted rule-following for judgement, so that his individual will left almost no trace on the massacre. Arendt's word "banality" describes an absence of thinking, not an ordinary personality. This is often read backwards: she was not excusing Eichmann but pointing out that the deepest evil need not depend on any depth.
Moravia's novel The Conformist renders this mechanism as an inner process. What the protagonist wants is neither power nor ideology but normality — to be accepted by the group and not to stand out. To preserve that normality he becomes an agent of the regime and ends up taking part in the murder of his former teacher. The novel's answer is cold: when normality itself becomes the goal, responsibility is handed over by the smoothest possible route, and the one handing it over does not even feel the transfer.
The counter-case: blaming "the system" is another form of exemption
At this point the argument can slide toward a more dangerous position: since responsibility is diluted by structure, the individual need not answer, and the system should. That inference has to be blocked, for three reasons.
First, it turns criticism itself into an exemption. If every executor can say "the system made me do it," then no one is answerable — and that is precisely the condition under which wrongdoing persists. In Responsibility and Judgment, Arendt draws a deliberate distinction between personal responsibility and collective guilt: the language of collective guilt sounds profound but in practice means that everyone bears a little less. Her insistence runs the other way — it is precisely because one is inside a structure that the obligation to judge must be retained. Postwar Germany offers the mirror image: Harald Jähner records in Aftermath that a society rebuilding order tends to do so alongside a systematic silence about the guilt just past. Silence is not denial; it is a more economical way of handling the matter, letting responsibility be neither assumed nor fully overturned.
Second, structure does not cancel choice; it only raises its cost. Intel could exit the memory business, IBM could open its interfaces, Bell Labs could diffuse its patents — each case shows that exit is possible and that the price is merely a price. Calling structure an irresistible force denies the choice space that genuinely exists inside it.
Third, the division of labour has real moral value, a point critics routinely overlook. It is what makes large-scale cooperation possible, and it also shields executors from being driven directly by the will of a single person. What Julien Benda attacks in The Treason of the Intellectuals is not specialization as such but the moment intellectuals voluntarily abandon universal values and enlist in partisan passions — the fault lies in the surrender of values, not in the division of labour. Research in the vein of How Groups Encourage Misbehavior asks the same question: how groups use role assignment and socialization to make individuals do what they would never do alone. A society with no division of labour at all would not be a moral paradise; it would be one in which everyone must personally execute everything. The problem is not the division of labour. It is that after the division, the paths of attribution were never rebuilt in step with it.
The judgement: answerability is a design problem, not nostalgia
Gathering these threads yields a frame more useful than "should the individual or the system answer." What needs repairing is not a return to a pre-division condition — that is neither possible nor desirable — but keeping responsibility traceable to specific persons after the division has happened. Three questions test an organization's answerability:
- Does the record of a decision land on a specific person? A document that says only "it was decided after study" is a device for evaporating responsibility, not a neutral piece of paperwork.
- Can those who execute know what their work points toward? If the outcome of a stage is entirely invisible to the person performing it, then "not knowing" was manufactured by the institution rather than being a personal lapse.
- Is dissent permitted and recorded? If the cost of voicing a reservation exceeds the cost of silence, then "no one objected at the time" constitutes no defence at all.
Finally the limits, stated plainly. First, answerability does not mean loading the responsibility for complex problems onto those at the far end of the chain — making frontline staff bear the full consequences of executive decisions is another injustice and destroys the willingness to cooperate. Second, in some domains strict proceduralization is itself a moral requirement: the value of medical protocols, flight checklists, and safety codes comes precisely from not leaving much room for on-the-spot individual judgement, and there optimality is a secondary variable. Third, tracing responsibility has costs, and an excessive culture of blame pushes people toward doing nothing — an organization in which no one dares to sign and one in which no one has to answer are two faces of the same problem.
Together these three point to one conclusion: responsibility is not a quantity that can be divided indefinitely; it has to be designed as a chain that can be traced. Division of labour split the work, and it split the blame — and that second cut was never properly accounted for.
