The Machine Without an Alibi: Technology, Algorithms, and Automated Cruelty

The previous entry ended where language begins to fail as a restraint: the decision survives, but the words that should expose its authorship can be hollowed out, inverted, or made to sound inevitable. Technology gives this evasion a new architecture. A judgment that once arrived as a sentence from a person can now arrive as a score, a flag, a ranking, a threshold, or a recommendation generated elsewhere. The result may be faster and more consistent. It may also be harder to answer.

Automation does not create arbitrary violence from nothing. Every technical system begins inside a prior world of categories, objectives, measurements, and permitted errors. Someone decides what will count as risk, success, fraud, danger, eligibility, productivity, or acceptable loss; someone selects the data through which those terms become visible; someone determines how much uncertainty the institution is willing to transfer to the people being classified. The code inherits these decisions even when the interface no longer shows them.

This is why technological mediation cannot be condemned simply because it is technological. Rules, statistical models, and automated procedures can reduce favoritism, expose inconsistency, and prevent the private whim of an official from becoming law. A machine may be less prejudiced than a person in one setting and more efficient at reproducing prejudice in another. The question is not whether calculation is humane or inhuman in itself, but what relation to judgment it creates.

Foucault’s analyses of apparatuses of security help clarify the shift. Modern power increasingly governs populations not by issuing one command to each body but by arranging fields of probability, circulation, normality, and risk (Security, Territory, Population). The technical apparatus does not merely punish after the fact; it sorts possibilities in advance. It asks which person resembles a danger, which pattern predicts default, which movement appears anomalous, which body belongs on which side of a threshold. At that point, explanation can become circular. The person is judged risky because the system assigns a high risk; the system is trusted because it processes more variables than any individual could see; and the difficulty of reconstructing the path from person to verdict is then presented as evidence of technical sophistication rather than as a problem of accountability. A decision remains, but its author has become increasingly faceless and difficult to locate.

Credit scoring, predictive classification, border screening, facial recognition, and other forms of automated sorting differ radically in purpose and consequence. Yet they can inflict immense material or symbolic violence. A crucial methodological principle follows: they should not be collapsed into a single indictment of “the algorithm.” What joins them analytically is narrower: each translates a singular person into a pattern that can be compared with other patterns, then uses that comparison to distribute access, scrutiny, delay, suspicion, or opportunity.

This new instrumental declension of the Logos—the logical reduction of singular being to a category, a numerical pattern, or an abstraction—is indispensable to administration. No institution can encounter millions of people as if every decision were a first encounter. But this reductive power of the Logos also creates a moral temptation: the digital proxy begins to stand in for the person. A location becomes intention; a correlation becomes character; a similarity becomes evidence; an error rate becomes acceptable because its victims are statistically dispersed.

The danger is not simply that a model may be wrong. Human judgment is wrong as well. The deeper danger appears when error becomes difficult to contest because the decision has been displaced into a technical process whose authority increases with its opacity. A person can challenge an official’s stated reason; it is harder to challenge a threshold no one in the room claims to have chosen. Levinas’s insistence on the face is useful here if it is not romanticized. The ethical face is not a guarantee that physical proximity produces compassion. It names the interruption of totalizing knowledge by a singular being who cannot be exhausted by the categories through which I know them (Totality and Infinity). Automated classification performs the opposite intellectual gesture: it must make the singular commensurable before it can act.

The issue is therefore not that data are unreal. Data can reveal patterns invisible to intuition and correct self-serving myths about supposedly neutral institutions. The problem begins when the data point becomes sufficient—when the representation ceases to be evidence about a person and becomes the person for the purposes of power. At that threshold, recognition becomes administration. The individual is no longer encountered as someone to whom a reason must be given, but processed as the bearer of a probability. And because probability has no face, the harm produced by its application can acquire the clean affect of an unprejudiced technical result.

Here the governing distinction of this project becomes non-negotiable. An algorithm cannot literally enjoy. It has no desire, shame, anxiety, sovereignty, unconscious, or jouissance. To say that code “wants” to exclude or that a machine “takes pleasure” in error would merely replace one mystification with another. The technical system executes operations. The libidinal economy belongs to the human and institutional world in which those operations acquire authority.

And so the Sadean question enters one level higher. What forms of enjoyment become possible when a person or institution can act through an apparatus that seems to decide by itself? There may be relief in surrendering judgment, prestige in possessing a system others cannot question, exhilaration in technical mastery, bureaucratic comfort in consistency, political profit in appearing severe, or simple convenience in transferring the cost of uncertainty to someone whose complaint will not interrupt the process.

Kafka is again immensely helpful here, as are Arendt’s historical reflections on bureaucratic evil. None of these enjoyments needs to amount to conscious sadism. That would repeat the conceptual mistake this project resisted in its treatment of bureaucracy. A systemic libidinal economy is not the secret psychology of every programmer, analyst, operator, clerk, or official. It names the structure in which exemption from responsibility, distance from consequence, and the pleasure of impersonal authority can become institutionally available even as agency grows increasingly disincarnate.

Consider the practice of war. Drone warfare makes the problem of mediated responsibility unusually stark, though technical mediation should not be confused with automation as such. A strike may involve intelligence, human authorization, rules of engagement, and multiple forms of review. Yet the screen, the sensor, the metadata trail, and the vocabulary of precision can place extraordinary distance between the person exposed to violence and those who authorize it. The danger lies not in distance alone, but in what distance can permit one side to elide altogether: an encounter with the embodied presence of the person subjected to violence.

Precision is morally valuable when it reduces harm. It becomes an alibi when technical accuracy is allowed to answer a different question: not whether the target was correctly identified according to the system’s criteria, but whether the criteria, threshold, objective, and permitted injury were themselves justifiable. The perfect execution of a procedure cannot redeem an indefensible procedure.

Technology therefore returns us to Derrida’s pharmakon. The same apparatus can discipline and poison, reveal and conceal, protect and expose (Dissemination). A model can uncover discriminatory patterns that ordinary non-digital judgment refuses to see; it can also stabilize those patterns by turning yesterday’s outcomes into tomorrow’s predictions. A surveillance system can locate a missing person or convert an entire population into permanently legible risk. The cure and the toxin share the same code.

For this reason, the ethical problem cannot be solved simply by demanding a human “in the loop,” as if human presence were automatically humane. A human can rubber-stamp a machine, while a machine can expose a human’s prejudice or hateful passions. What matters is whether the architecture makes judgment contestable through mechanisms of checks and balances: whether reasons can be reconstructed, classifications challenged, errors repaired, and the person affected allowed to appear as more than the data assigned to them.

Responsibility must therefore travel along the entire technical chain. It belongs to the institution that commissions the system, the officials who define its purpose, the designers who choose its proxies and thresholds, the managers who decide what error is acceptable, the operators who treat its output as command, and the political order that determines who bears the consequences. Digital distribution and delegation make responsibility complex, but not ownerless.

The most seductive alibi is neutrality. “The system flagged you” sounds different from “we decided to treat this pattern as sufficient reason to burden you.” The first sentence converts an institutional choice into a natural event. The second restores the missing subject. The ethical task begins in that restoration of human agency: to force every technical necessity back into the grammar of decision and accountability. One condition of an ethical stance is the constant reminder that the machine is not outside the libidinal economy. Its strategic uses are shaped by the economies that pay for it—market forces that can reward subjection and control through speed, prediction, surveillance, and the capture of attention—and by institutions that discover new value in reducing singular beings and becomings to data. Once attention itself becomes measurable, sortable, and sellable, the technical apparatus opens directly onto another libidinal economy: not merely the administration of risk but the production and capture of objectified desire itself—the urge to have and possess.

This is where the next entry begins. The algorithm does not enjoy, but the economy in which it operates—its articulation around profitable circuits of repetition, obsolescence, waste, and disposability—is impossible to comprehend without a diagnosis of the libidinal economy of consumer jouissance. The machine has no alibi because it has no moral agency to begin with. The alibi belongs to us—even as we hide behind it.