Data for: Judgment Cost Theory: The Accountable Bearing of Judgment in an Age of Cheap Prediction

<p>Replication data and analysis code for <em>Judgment Cost Theory: The Accountable Bearing of Judgment in an Age of Cheap Prediction</em> (West &amp; Cosimano, 2026; SSRN preprint: <a href="https://ssrn.com/abstract=6992679">https://ssrn.com/abstract=6992679</a>). The study develops and validates the Judgment Cost Index (JCI), an occupational measure of judgment cost &mdash; the burden of making and owning an accountable decision under irreducible uncertainty &mdash; constructed from ten O*NET 29.3 descriptors spanning two facets, accountability for outcomes and irreducible uncertainty. The index is related to nine independent measures of occupational AI exposure across 763 six-digit SOC-2018 occupations in hierarchical regressions controlling for cognitive complexity, routineness, education, and wage. Across eight of the nine measures, including all five generative-AI-era indices, higher judgment cost predicts significantly lower AI exposure beyond the controls; the result survives an eight-part robustness battery, 930 alternative constructions of the index, and factor-score weighting.</p> <p>This deposit contains the redistributable analysis dataset (constructed index, sub-scores, controls, employment, and wages for 763 occupations), item-level z-scores for all index and control items, a complete codebook with O*NET element IDs, a step-by-step replication protocol with data-source URLs and versions, the sole verified analysis implementation (Python; pandas and numpy only), summary outputs for the specification-curve, item-influence, factor-weighting, and path-decomposition analyses, and publication-resolution figures. Nine third-party AI-exposure measures are excluded because redistribution rights rest with their original authors; the replication protocol documents exactly where to obtain each and how to merge it, and the reconstruction reproduces the archived variables at r&nbsp;=&nbsp;1.000.</p> <p>Data and documents are released under <a href="https://creativecommons.org/licenses/by/4.0/">CC BY 4.0</a>; code under the MIT License. Includes information from the <a href="https://www.onetcenter.org/database.html">O*NET 29.3 Database</a> by the U.S. Department of Labor, Employment and Training Administration (USDOL/ETA), used under CC BY 4.0; O*NET&reg; is a trademark of USDOL/ETA. Employment and wage figures are from the U.S. Bureau of Labor Statistics <a href="https://www.bls.gov/oes/">Occupational Employment and Wage Statistics</a>, in the public domain.</p>