Override Log — Reviewer Package

Abstract

Reviewer package accompanying a manuscript submitted for double-blind peer review.The manuscript proposes a five-element record of each human adjudication of a machine-generated classification in AI-assisted spatial analysis workflows: decision point, machine proposal, human disposition (affirm / reject / modify), stated justification, and downstream consequence. This deposit contains the record itself rather than a description of it, so that the paper’s central claim — that such adjudications should be open to inspection — can be checked against the artifact.Contents:schema.json — the five-element specification, with the determination each element supports and the design requirements for append-only integrity, human-readable primacy, and coupling to lock events.blank_template.csv — an empty implementation template for reuse.trace_A.json — a worked adjudication from a documentary registry pipeline, in which a human rejects a generative model’s exclusion verdict.trace_B.json — a worked adjudication from a rule-based pipeline, in which a human modifies a pattern-matched classification.negative_case.json — an unlogged exclusion, included to show what the record makes visible by contrast.README.md — field definitions and reuse notes.The two traces come from pipelines whose proposers differ in kind — one a stochastic generative model, one a deterministic rule set — so the schema is demonstrated across both.