Technological change does not displace workers. Policy does. The machine is the occasion; the distribution of its consequences is a decision made by people with the authority to make different decisions. Keeping these two things confused is not an accident — it is how the decision gets made without appearing to have been made at all.
The standard account runs like this: technology advances, certain jobs become obsolete, workers must adapt. The passive construction is doing enormous work here. Jobs do not become obsolete. Firms eliminate them because the technology makes elimination profitable, and the regulatory and tax environment makes it easier to eliminate the job than to retrain the person. The inevitability is constructed. It is built from specific choices about corporate tax treatment of automation, about the absence of adjustment obligations on firms receiving productivity gains, about the deliberate weakness of collective bargaining rights that might otherwise force negotiation over the pace of change.
What we owe displaced workers, then, cannot be answered without first answering who chose to make them displaced, and under what conditions that choice was permitted.
The instinct of liberal democracies has been income replacement — unemployment insurance, retraining programs, extended benefits in particularly affected sectors. These are not nothing. But they share a structural assumption that reveals itself under pressure: they treat displacement as a weather event and the worker as its victim, rather than treating displacement as the output of a system and asking who the system serves. Retraining programs, specifically, have a particularly poor empirical record — the evidence from Trade Adjustment Assistance in the United States, evaluated repeatedly over two decades, shows minimal wage recovery for displaced manufacturing workers. We fund the programs. They do not work. We continue funding them. This is not confusion. It is the function of appearing to respond without disturbing the distribution.
Apply Kant's test here. If you were a fifty-three-year-old logistics worker in Leipzig whose job was eliminated when the warehouse automated, and someone in a position of institutional power told you that the appropriate response was a retraining voucher and three months of income support — would you find this adequate? Not as a matter of preference but as a matter of principle: would you accept that this obligation was discharged? The question answers itself. The people designing these programs would not accept this were it applied to them. They do not design it for themselves.
The harder claim is this: when a firm captures the productivity gains of automation, it has taken something from a system of social cooperation that made that productivity gain possible. The workforce, the infrastructure, the legal and financial architecture, the decades of institutional knowledge embedded in the labor it is now discarding — none of this is the firm's private creation. The extraction of value from social cooperation while the costs of reorganizing that cooperation are externalized onto individuals is not a market outcome. It is a political settlement. It can be renegotiated.
Renegotiation looks like this: an obligation, legally defined, on firms above a certain scale that automation events trigger — not suggestions, not tax incentives, but binding requirements for transition periods, wage maintenance, and genuine retraining with demonstrated outcome accountability. It looks like taxing the productivity gains of automation directly and redistributing them through a social wealth fund rather than letting them concentrate at the ownership layer. It looks like recognizing that when you have broken the implicit contract of employment as a stable social institution, you owe the people that institution organized their lives around something more than its wreckage.
None of this requires the claim that automation is bad or should be stopped. Simone Weil wrote about affliction — the specific damage done to human beings not merely by suffering but by suffering that makes them feel that their suffering does not matter, that they are invisible to the social order they believed themselves to belong to. The displaced worker who has followed every instruction, done everything asked, found themselves eliminated anyway, and been handed a pamphlet about coding bootcamps — this person is experiencing something specific. The harm is not only material. It is the discovery that the social contract was not mutual, that the obligations ran only one way, that the institution they served did not consider itself as having served them.
We do not discharge a debt by acknowledging that the debt exists. Acknowledgment is the beginning of the question, not the answer to it.
The political economy of automation will continue producing displacement on a significant scale. Robotics in logistics, AI in professional services, autonomous systems in transport — the sectors and the workers and the timelines are not abstractions. The people who will bear these costs are not hypothetical. They are specific, countable, locatable. The question of what is owed to them is not a philosophical puzzle to be held in suspension while the policy process runs. It is a choice being made, right now, by its ongoing refusal to be made.
That refusal is itself a policy choice. It benefits someone. The analysis of who it benefits is where serious engagement with this question has to begin.