The Decoupling of Productivity from Human Hands: Why AI Breaks the Abstraction Ladder

The optimistic case for AI and work leans on history: mechanisation emptied the fields and filled the factories; automation emptied the factories and filled the offices. Each time, productivity gains created new, more abstract categories of human work, and displaced labour climbed the ladder. The paper takes that history seriously enough to ask what actually made the climbing possible — and finds a condition, not a law. Workers were re-absorbed because each new rung consisted of work that only humans could do. The machine that displaced them could not follow them up.

AI is the first general technology that climbs. Whatever new category of cognitive work emerges above the current rung, a system trained on human output can, with a lag measured in years rather than generations, perform it more cheaply. The absorption mechanism that turned two centuries of displacement into two centuries of redeployment stops operating, and the productivity returns that used to fund a new tier of wages accrue to compute instead.

The paper grounds the mechanism in the economics of structural change and task displacement — Pasinetti’s structural dynamics, the Autor lineage on task polarisation, Stansbury and Summers on declining worker power — and tests the failure mode against the United Kingdom’s coalfields: a profession that shed over two hundred thousand jobs in a decade, where reskilling programmes and a billion pounds of regeneration did not restore the communities, and displacement hid for decades in incapacity rolls. That is what mass mitigation looks like when the ladder is gone locally. The question the paper leaves for policy is what it looks like when the ladder is gone everywhere — the question the Cultural Alignment paper takes up.