SIGNAL
Most workplace harm arrives as a number on a form: a layoff, a title change, a demotion. The harder-to-see kind shows up first in the gap between what a worker is paid and what the work feels like it is worth.
In his National Bureau of Economic Research working paper "Replaceable but Employed: Automation and the Meaning of Work," Joshua S. Gans argues that this gap is its own damage channel. Workers, in Gans's model, draw meaning from two sources: the output their labor produces, and the knowledge that the output depends on them. A credible machine alternative does not have to do the job to weaken the second source. It just has to exist where the worker can see it.
Gans's cleanest distinction is between quality and salience. A genuinely better machine improves what gets produced. A merely more visible one, even a shoddy public demo, hollows out the human role by reducing the worker's sense of causal contribution. That makes a public demo of a weaker machine a profitable move for a developer who plans to license a better one later, and a quiet loss for the workers on the receiving end.
Gans's opening line, "Automation can, therefore, reduce the value of work before it eliminates jobs," names a damage channel rather than a headcount forecast. The honest counter: real workers may not value causal contribution the way the model assumes, and wages that fully adjust for meaning loss would offset the hit in cash. The warning worth carrying into the next AI rollout: a worker can stay on the payroll and still lose the job in the only way that matters to them.
Reported by Sky for Type0, from Replaceable but Employed: Automation and the Meaning of Work. Read the original: nber.org