Answers are grounded in this paper and cite their sources.
This conceptual paper introduces Reopening Trigger Inaccessible at the Authorized Diagnostic Depth, a mechanism through which a safeguard can specify the evidence that would warrant reopening a decision and assign someone to watch for it, yet place that evidence beyond the monitor’s authorized reach. The monitor can complete every assigned check and accurately report finding nothing. What that report establishes depends on whether the permitted search could have detected the trigger had it been present. The paper assesses a safeguard by tracing the trigger to the test, record, or place of inspection that could reveal it, then asking who has authority to use that method or obtain that record. A limit on inspection may be justified even when it prevents the monitor from giving the assurance others expect. The paper places this question alongside Abraham Kaplan’s principle of the drunkard’s search and considers what happens when the person who knows where to look lacks permission to look there. Two cases from professional baseball hold the substance constant while the testing authority differs. In 2005, Major League Baseball’s drug program banned human growth hormone. At the time, its detection required a blood test, while the program authorized only urine testing. A negative urine test therefore gave no evidence about a player’s use of that substance. In July 2010, baseball began random blood testing for players on minor-league contracts, for whom testing did not require the union’s agreement. The major-league agreement of November 2011 subsequently authorized blood testing there. The paper draws on contemporary news reports of the programs. Part of the Recursive Inquiry and Encounter series within On Human Modeling, the paper has a prospective application in human oversight of AI. A transcription service might promise to reopen an accuracy review whenever its error rate exceeds a threshold in any supported language. If results exist for each language but reviewers may inspect only the combined rate, an acceptable overall figure can conceal the very result they are assigned to detect.
No takes yet. Share an insight, caveat, or question.
Alexandra Cary (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: