From Formal Authority to Practical Human Control: A traceable method for reconstructing human control in automated decisions

Institutions often treat a named reviewer or approval step as evidence that an automated system remains under human control. That inference can fail when the person lacks timely information, comprehension, authority, an opportunity to challenge the system, or the ability to affect the outcome. This paper develops a frozen, evidence-traceable procedure for representing formal authority, practical control, and unresolved evidence in bounded public incident records. Before case screening, the protocol fixed the evidence cutoff, candidate collections, eligibility rules, screening order, selection strata, and missing-data treatment. Three versioned historical case packets record claim-level provenance, categorical evidence states, and a chain from information access through effect. A frozen reanalysis changed six Oko pre-action stages from supported to partially supported. The current assessments record partial support across those Oko stages, supported authority with unsupported practical conditions in Patriot ZG710, and supported authority with unresolved conditions in the Patriot F/A-18C case. An open-index search produced 2,431 deduplicated records. Author screening and forward-citation review yielded five bounded source propositions for manuscript use, two background-only sources, and six quarantined sources. The contribution is a governance-specific integration with a visible correction path. Institutions claiming human control bear an ethical burden to preserve evidence of practical force because unsupported attribution can obscure responsibility and exposure to harm. Three retrospective, single-assessor cases supply no population estimate, causal validation, independent-reliability result, or transfer claim for current learned systems. Status: Preprint v0.14.0. This manuscript has not undergone peer review. AI-assistance disclosure: Generative AI tools assisted with literature discovery, metadata organization, preliminary screening proposals, decision-ledger preparation, draft development, code generation, figure and consistency checks, and language revision. Mark Julius Banasihan defined the research question and protocols, reviewed and approved the 89 screening decisions, approved the evidence and assessment decisions, verified the sources used for claims, controlled the repository releases, and remains responsible for the analysis, interpretations, errors, and submitted text.