Simone Weil, writing in London in 1942 and 1943 in the essays gathered after her death as Attente de Dieu, insisted on a distinction that ordinary moral language collapses. There is suffering, souffrance, which is pain, grief, illness, loss, and which the soul survives and sometimes ripens under. And there is malheur, affliction, which she defines by three simultaneous components: physical pain or its equivalent, social degradation, and a mark left on the soul that persists after the occasioning cause is removed. Affliction, she writes, takes possession of the soul and brands it to its depths with a slave's mark; it is not proportional to injustice, it does not select the guilty, and its most characteristic property is that it makes the afflicted complicit in their own contempt. The afflicted come to believe the verdict passed on them. Weil's factory year at Alsthom and Renault in 1934 and 1935, recorded in the Journal d'usine, was the empirical ground: she went in a philosopher and came out, in her own account, marked, unable to think of herself as anything but a slave, and she did not recover the capacity to consider herself a person with rights.

I want to argue that the current deployment of statistical models across the trades is producing affliction in Weil's technical sense rather than suffering in the ordinary sense, and that almost the entire commentary on the subject, whether it counts jobs saved or jobs lost, is measuring souffrance and therefore measuring the wrong thing. The wage is a measure of suffering. The mark is not on the wage.

Begin with what Weil identified as the first and least discussed of the three components: uprooting. In L'Enracinement, written in the same final months, she names déracinement the disease of the age, and she means something precise by rootedness: participation, real and active, in a collectivity that preserves treasures of the past and expectations of the future. A trade is such a collectivity. A radiographer belongs not to a hospital but to radiography, to a lineage running from Röntgen through a century of accumulated seeing, with its own criteria of excellence, its own scandals, its own dead. This is what Alasdair MacIntyre in After Virtue (1981) formalised as a practice: an activity with goods internal to it, standards of excellence partly definitive of it, and a history through which those standards extend. Goods internal to the practice are the ones you cannot get by any other route, and they are the roots.

What model deployment does, and this is the descriptive core of the mutation, is sever the internal good from the external one while leaving the job title intact. The radiologist reading behind a detection model still earns, still holds the licence, still signs. But the sequence in which judgement was formed, the long apprenticeship of seeing a thousand unremarkable chests so that the thousand and first declares itself, is now performed upstream by a system that does not teach. The junior who once acquired the eye by carrying the volume now adjudicates flagged cases, and adjudication is not how the eye is grown. I have called this elsewhere the hollowing of the middle rungs; here I want a more exact term. Call it *practice severance*: the condition in which a trade retains its external goods, its pay, its credential, its social recognition, while the internal goods that made it a practice are transferred to a system that cannot transmit them to a successor. Practice severance is uprooting with the payroll left running, and this is precisely why it does not register as suffering. Nobody is hurt. Everybody is marked.

The second component, social degradation, arrives by a route that is easy to misread as improvement. Weil noticed at Renault that the humiliation of the factory was not primarily the foreman's contempt but the structural fact that the machine set the pace and the worker's body had to become its residue: she describes losing the sense of having a right to be there, of being a rebut, a reject, tolerated. The current instrument of degradation is measurement of a novel kind. Warehouse telemetry, the pick rate compared against an algorithmically derived expected rate; the call centre sentiment score; the developer whose contribution is now legible as the ratio of accepted machine suggestions to keystrokes typed. What is new is not surveillance, which is ancient, nor the rate, which Taylor systematised in The Principles of Scientific Management (1911) with the stopwatch and the differential piece rate. What is new is that the standard is no longer a negotiated fiction about a human body but the output of a system that does not tire, and against which the worker's shortfall is not a bargaining position but an ontological fact. Taylor's rate could be disputed because a man with a stopwatch had set it and another man could time it differently. A model's expected rate is a distribution; you do not argue with a distribution. The result is a species of degradation that generates no grievance, because grievance requires a defendant.

Here is the raid I want to make. Max Weber, in the closing pages of Die protestantische Ethik, describes the stahlhartes Gehäuse, the shell hard as steel into which the rationalised order encloses those born into its mechanism, and he is right about the enclosure but wrong, I think, about the temperature. The enclosure Weber describes is cold and impersonal and therefore leaves the inhabitant's self-regard intact; you can hate a bureaucracy and remain a person. Weil's insight, developed against the same material a generation later, is that when rationalisation reaches the point where the criterion of adequacy exceeds any human performance, the enclosure stops being external and is internalised as a verdict on the self. Jacques Ellul in La Technique ou l'enjeu du siècle (1954) named the mechanism, the autonomous self-augmentation of technique in which each solution generates the problem requiring the next, but Ellul writes as a diagnostician of systems and has almost nothing to say about what happens inside the person who fails the system's standard. That interior is Weil's subject and it is the one currently unattended.

The third component is the mark, and the mark is the strangest and most verifiable claim in Weil's account. She observes that affliction produces in the afflicted an agreement with the judgement against them, a self-contempt that survives every rational demonstration of its groundlessness, and that this is why the afflicted are avoided: their presence is contagious, and the healthy sense it. Test this against a translator. Machine output has not made translation impossible; it has made the translator's rate a matter for post-editing, in which the human is paid a reduced fee to repair a draft they did not write and would not have written. Ask what happens to the translator's estimate of the twenty years spent acquiring the two languages. The economic fact, a lower rate, is suffering. The mark is the moment the translator concedes, privately and without argument, that the twenty years were a mispricing of their own life, that the accumulated feel for register and cadence was always a redundancy waiting to be discovered. That concession is the slave's brand. It does not appear in any labour statistic and it does not reverse when the rate recovers.

Call this *retroactive devaluation: the mechanism by which a system's competence in a domain reaches backwards through a life and reassigns the meaning of the years spent becoming competent in it. This is the specific cruelty of the current deployment and it is what distinguishes it from prior mechanisation. The handloom weaver displaced by the power loom lost a livelihood and knew whom to break; the loom did not tell him his skill had never been skill. E. P. Thompson in The Making of the English Working Class* (1963) documents a displacement furious with dignity, and fury is a sign of intact selfhood. Retroactive devaluation produces no fury. It produces the quiet reclassification of one's own biography as an error, which is exactly the condition Weil describes when she says the afflicted lose the capacity to believe they deserve anything at all.

Weil offers one reversal and only one, and it is worth taking seriously precisely because it is so unpromising as policy. Attention. Not sympathy, not activism, not redistribution, though she supported redistribution: attention, defined in the essay on school studies as the suspension of thought, leaving it available, empty, penetrable by the object; the capacity to look at an afflicted person and ask, in effect, what are you going through, and to wait for the answer without supplying it. She insists this is nearly impossible, that it is a miracle, and that the afflicted are almost never granted it, because affliction is boring and the afflicted are unpleasant company and the healthy have every reason to look away. Her claim is that this attention, and nothing else, unbrands the soul, because the brand consists in the belief that one is nothing, and being genuinely attended to is the only evidence against it that the afflicted cannot dismiss.

What does this mean for the trades under severance? It means the interventions currently discussed are all souffrance interventions, addressed to the wage: retraining budgets, transition allowances, income floors. Each is defensible and none touches the mark, because the mark is not a shortfall of money but a shortfall of witness. The functional equivalent of Weil's attention, in a firm deploying models across a craft, would be an institution that does what no severance package does: registers, in a form the practitioner can see, that the accumulated judgement of the trade was real, is real, and is being taken from the trade for reasons of cost rather than reasons of truth. That distinction, cost against truth, is the whole of it. A radiographer told the model is better can be marked. A radiographer told the model is cheaper, and shown the arithmetic, cannot be, because the verdict has been returned to the domain of decisions taken by people, where it can be resented, contested, and survived.

This is not a consolation and it is not a programme. It is a specification of what the damage actually is, offered against a discourse that has settled on employment counts because employment counts are countable. Weil's contribution to the analysis of technique is that she went into the factory and found the injury below the injury: not the loss of the wage but the loss of the standing to want one. The models now entering law, medicine, translation, radiography, freight and code are performing that operation at a scale and speed the assembly line never approached, and they are performing it on the credentialed and the innocent alike, which is Weil's whole point about affliction: it does not select for guilt, and it has never had to explain itself.

The finding, stated so it can be built on: automation's characteristic injury is not unemployment but decertification of a life, and the only known reversal is a witness who names the reason as cheapness and refuses to name it as truth.