Execution and Evaluation: A New Occupational Measure and Long-Run Employment Gradients

URL SCAN: Execution and Evaluation: A New Occupational Measure and Long-Run Employment Gradients
FIRST LINE: Economics > General Economics

The Dissection

This is a measurement paper, not an AI-displacement autopsy. It scores 19,265 O*NET task statements, separates execution from evaluation, tests reproducibility, and maps employment gradients.

Its result is narrow but significant: execution-heavy white-collar occupations have weaker employment growth in every window since 2012. The gradient is secular, not demonstrably an AI-era event. The post-2022 capability steepening is dated but explicitly not causal.

The paper’s real function is to replace the crude phrase “AI exposure” with a task-function distinction. It also blocks the lazy claim that every post-2022 labor-market movement is caused by AI.

The Core Fallacy

The central conceptual weakness is category substitution: occupational exposure and employment growth are treated as proxies for productive necessity.

The paper does not prove that execution-heavy workers have become economically unnecessary. It shows only that their occupational employment growth is weaker. That is a gradient, not proof of P3—the collapse of mass productive participation.

The execution-evaluation distinction is also a lag marker, not a permanent moat. Evaluation may currently be harder, but the abstract supplies no reason to believe it is structurally immune to automation, concentration, or machine-assisted replacement. The bottleneck is simply the next target.

Its causal restraint is methodologically defensible. But “not causally attributable” is not a structural explanation. It is a boundary around what this dataset can prove.

Hidden Assumptions

• Execution and evaluation are separable, stable functions rather than shifting parts of the same workflow.
• Evaluation remains human-essential instead of becoming automatable or concentrated among a smaller number of owners.
• O*NET task statements and model-based scores adequately represent actual work, deployment, cost, and performance.
• Employment growth measures productive participation rather than institutional lag, occupational relabeling, demand changes, or labor hoarding.
• A non-rejected difference in slopes can be treated as evidence of equality. It cannot; non-rejection is not proof of equality.
• Between-family employment gradients reveal the underlying mechanism rather than merely documenting where existing restructuring has already concentrated.
• The 2012-to-present gradient can be interpreted without separating AI from other technological and organizational forces.

Social Function

Classification: partial truth, prestige signaling, and transition management.

It is partial truth because it identifies a real asymmetry: producing output is being commoditized faster than judging it. It is prestige signaling because formal rubrics, reproducibility, and occupational scoring convert a destabilizing social process into an administratively respectable measurement problem. It is transition management because institutions can catalog the erosion as a gradient while avoiding the harder questions of ownership, bargaining power, and who remains economically necessary.

That function need not be intentional. The effect is the same: the system receives a better thermometer and another excuse not to discuss the patient’s terminal condition.

The Verdict

This is a useful instrument and an incomplete diagnosis. It supports the DT claim that execution is exposed before evaluation and that labor-market erosion predates the convenient 2022 timestamp. It does not establish that AI caused the gradient, that evaluation is a durable human refuge, or that mass productive participation survives.

Under the hardened framework: P1 is compatible but unproven; P2 is untouched; P3 is not measured. The paper is therefore a lag map, not a survival map. Anyone using it to declare either “AI has already destroyed employment” or “AI is not destroying employment” is converting a cautious empirical paper into ideological ammunition.