Loading...
Loading...
Contributors:
Dataset URI:
10.5281/zenodo.22867822Collective AI procedure and authority over training on worker-created material are distinct contractual allocations, even within the same agreement. This empirical study examines a source-defined frame of 46 published Writers Guild of America East directory entries, comprising 53 linked instruments across online media, broadcast news, nonfiction television and podcasts. Forty instruments are readable, yielding 34 complete agreement bundles and twelve unresolved bundles. Among the 34 complete bundles, sixteen provide AI-specific collective procedure without express training control, one provides training control without the specified collective procedure, one contains both, and sixteen contain neither. Collective procedure therefore appears in seventeen bundles and express training control in two: a within-bundle difference of 44.1 percentage points. Assigning every unresolved bundle training control without procedure still leaves a positive full-frame difference of 6.5 percentage points. Missing sources alone therefore cannot reverse the direction of the principal finding at the observed classifications; sensitivity to disputed classifications is assessed separately. The analysis distinguishes collective procedure, triggered process, effects bargaining, training-input control, individual replica consent, participatory deployment approval and employer permission for employee tool use. The seventeen procedural arrangements are further disaggregated into six new-use notice arrangements with separately coded effects bargaining, nine new-use notice arrangements without that separately coded feature, and two policy-notification or working-group arrangements. Source-to-classification exhibits show why discussion is not approval, a mixed committee is not a unilateral union veto, and permission for an employee to use a tool is not permission for an employer to train on employee work. Alternative definitions, document-ambiguity exclusions, sector leave-outs and grouped-unit comparisons test the stability of the findings. The contribution is an empirical account of selective contractual control: negotiated procedure over implementation does not imply authority over every use of worker-created material. Within-bundle comparisons establish the configuration, while finite-frame missingness bounds identify which contrasts survive every possible completion of unresolved records. The results concern allocations in published agreement bundles, not worker coverage, national bargaining prevalence, realised deployment or welfare. An “AI clause” does not identify a common decision-maker, permission direction or form of control. The accompanying research materials provide the manuscript, online appendix, source-to-classification exhibit, frozen analytical data, source ledger, coding documentation, executable analysis and LaTeX sources. Article and page locators preserve the connection between the classifications and the underlying contractual provisions, including material conditions, exceptions and remedial limits. The replication workflow reproduces the numerical tables, sensitivity analyses and analytical figures from the frozen inputs.
Keyword(s):
MeSH Term(s):
Grants and Funding:
No Grants or Funding Data Available
Associated Publication(s):
No Associated Publication Data Available
Supplementary Data:
No Supplementary Data Available
Related Dataset(s):
No Related Dataset Data Available
The National Library of Medicine (NLM) provides information about biomedical datasets. Inclusion on an NLM Web page does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: NLM Web Policies (nih.gov)