Replication package for "Institutions as Correction Paths: Independence, Competence, and Answerability in the Governance of Artificial Intelligence"

Replication package for the manuscript "Institutions as Correction Paths: Independence, Competence, and Answerability in the Governance of Artificial Intelligence" (submitted to the Journal of Institutional Economics). Contents: coding protocol; case-selection criteria; the reconciled seven-capacity correction-path matrix for six documented warning episodes (codes D / D* / P / A / N/A); codebook; cell-level evidence table with primary-source pinpoints; observation-window dates, including the right-censoring of the marking episode; a reconciliation-and-limitation note; the fixed instructions, blank coding and justification templates, and agreement-computation script for the pre-specified independent-coding step; and SHA-256 checksums. All case material derives from public records; the evidence table records the citation level (page, paragraph, exhibit/heading, or document) for every coded cell.