What AI job loss predictions get wrong about technological change – e61 INSTITUTE

URL SCAN: What AI job loss predictions get wrong about technological change – e61 Institute
FIRST LINE: Predictions of imminent AI-driven job loss have been mounting.

THE DISSECTION

This is a textbook case of institutional lullaby — a competent-sounding economic reframe that smuggles in comforting historical analogies to neutralize a structural threat the author is intellectually obligated to acknowledge but politically incentivized to defuse. The e61 Institute is a policy-oriented think tank, which tells you exactly what social function this piece serves.

The author constructs a three-channel framework (automation, productivity, new tasks) that is technically correct as a description of past technological change. He then deploys this framework to argue the future is uncertain and the net effect of AI on employment is unclear. This is intellectual sleight of hand: using a historically accurate model of previous automation waves to manage anxiety about a qualitatively different structural rupture.

THE CORE FALLACY

The fundamental error is ** category confusion between prior automation and AI-driven automation**. The entire three-channel framework assumes a world where:

• Automation replaces specific physical or routine cognitive tasks
• Productivity gains create demand for complementary human labor
• New tasks emerge that humans can perform

The DT framework exposes why this is broken. Prior automation replaced specific tasks while leaving the human employment-wage-consumption circuit intact. AI is different in kind, not degree: it severs the mass employment -> wage -> consumption nexus itself because it can perform general cognitive labor at scale and cost. The author treats AI as another wave of automation when it is actually the automation of the conditions that made prior automation manageable.

The computer analogy is the most egregious error. Computers automated calculation but created software developers, cybersecurity experts, and e-commerce workers — all of whom became consumers and taxpayers and wage earners sustaining the system. AI automates the creation of those roles. There is no new task category analogous to "software developer" that only humans can perform and that absorbs mass numbers of displaced cognitive workers. The author offers "care, education, and hospitality" as the supposed refuge — sectors characterized by precisely the low wages, casualization, and inability to absorb mass unemployment flows that make this suggestion laughable on any serious scale calculation.

HIDDEN ASSUMPTIONS

1. Labor demand remains intact at the aggregate level — Assumed, not argued. The entire historical precedent argument depends on this, but this is precisely what P1 (Cognitive Automation Dominance) falsifies for the current wave.

2. New tasks will emerge at a rate that absorbs displaced workers — Assumes new task creation is endogenous to automation and proceeds at a rate comparable to displacement. No mechanism is offered for this; it is simply asserted based on historical pattern matching.

3. Productivity gains translate to expanded output and new labor demand — This assumes the Jevons Paradox dynamic works for AI. It ignores that productivity gains in AI don't require additional human labor inputs — that is the entire structural rupture.

4. Sectoral variation will cushion aggregate effects — "IT grew while agriculture shrank." True, but this assumes there are sectors with high income elasticity and positive demand response to productivity gains. The author suggests "care, education, and hospitality." These sectors have structurally constrained income elasticities. If anything, widespread AI-driven unemployment would reduce demand for these services among the displaced.

5. "So-so automation" is the worst case — The author treats rapid automation outpacing productivity as the danger scenario. He frames this as uncertain. Under the DT framework, this is not an alternative scenario — it is the default outcome. AI automating tasks faster than productivity gains can reallocate workers to new complementary roles is the mathematical consequence of AI having no marginal cost of reproduction.

SOCIAL FUNCTION

Transition management / elite self-exoneration / policy calming agent. This piece is precisely calibrated to:
- Acknowledge AI is genuinely disruptive (to appear honest)
- Reassure that the system has handled disruptions before (to prevent panic)
- Deflect toward income support policy tweaks (to preserve institutional relevance)
- Never name the structural discontinuity at the center of the thesis

The final paragraph is the tell. "Getting the policy foundations right matters not because we can predict what's coming, but precisely because we can't" — this is the language of someone who has concluded the outcome is unpredictable in order not to say it is structurally determined and terminal. It is epistemic surrender dressed as humility.

THE VERDICT

This piece performs the function of a sedative: it is accurate about the past, misleading about the present, and silent about the mechanism that makes the future categorically different. The three-channel framework is not wrong — it is simply inapplicable to a technology that automates cognition itself, at scale, without proportional new human task creation. The author offers historical reassurance while writing about a discontinuity. That is not analysis. That is institutional management of a crisis the institution cannot solve.