Candidate Exclusion Before Answer Selection: How an Inquiry Can Rank Its Candidates Correctly and Still Miss the Answer

This conceptual paper introduces Candidate Exclusion Before Answer Selection, the operation by which candidate generation leaves out a relevant possibility, placing it beyond the reach of ranking within the resulting set. Selecting an omitted candidate requires expanding that set through renewed generation or another source. The paper assesses inquiry through both the adequacy of the candidate set and the accuracy of ranking. A candidate considered and rejected is different from one never included. Poor fit or unexplained cases can prompt expansion, but accurate ranking alone does not establish that the relevant possibilities were available for selection. The discussion relates the operation to Stanford’s problem of unconceived alternatives and uses the identification of Helicobacter pylori as a cause of peptic ulcer disease as an illustration. The Nobel Assembly’s historical summary does not identify a particular inquiry’s candidate set or establish that bacterial causation was excluded from it. Documenting the operation requires evidence of the relevant omission. Part of the Possibility and Counterfactual Structure series within On Human Modeling, the paper offers questions about candidate sources, coverage, and control over admissible alternatives. Its framework can inform human–AI research on whether retrieval and candidate-generation processes include relevant possibilities before answers are ranked or selected.