Answers are grounded in this paper and cite their sources.
This conceptual paper introduces Depth-Gated Evidence Availability, a mechanism through which the scope of a diagnosis determines what evidence an inquiry can obtain. An investigator who treats a model’s assumptions as settled can test different values within it while leaving those assumptions unexamined. This creates a particular difficulty when the evidence that would justify widening the inquiry depends on a comparison that the current inquiry has no occasion to make. The paper assesses a choice of diagnostic depth by asking what questions the current frame permits, what evidence would bear on a deeper alternative, and whether the present investigation could obtain it. To identify the mechanism, one must trace a specific gap in the evidence to a question the frame excludes and show how asking that question makes the evidence obtainable. A narrow investigation can still be proportionate to its purpose. The value and cost of the missing comparison matter when deciding whether to pursue it. The paper places the mechanism alongside Thomas Kuhn’s account of how paradigms guide the selection of scientific problems and examines two historical cases. In 1901, the Wright brothers’ glider produced about one-third of the lift they expected. Their wind tunnel allowed them to compare different wing shapes under the same conditions. The comparison supported Lilienthal’s measurements for his own wing and exposed a problem in applying them to the Wrights’ wings. In 1909, Hans Geiger and Ernest Marsden asked whether alpha particles could return from a metal surface. A screen shielded from direct particles let them observe returns that earlier forward measurements could not register. Those observations later informed Ernest Rutherford’s proposal of a concentrated central charge. Part of the Recursive Inquiry and Encounter series within On Human Modeling, the paper has a potential application in human–AI evaluation. Reviewers who vary a prompt and score each answer may take for granted that the assistant received the relevant source passage. Asking which passages actually reached the assistant calls for a different record. That record lets reviewers distinguish a passage the system omitted from one it received and failed to use.
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: