Historical Non-Identification From Identical Records: Why a Rule That Reads Only a Record Cannot Guarantee Which History Produced It

This conceptual paper introduces Historical Non-Identification From Identical Records, the structural constraint that a rule cannot guarantee which history occurred when two admissible histories produce the same complete record available to it. The claim applies to a fixed information set and background. The paper distinguishes guaranteed discrimination from probabilistic inference, whose reliability depends on assumptions about competing histories. New independent evidence changes the information problem and may make discrimination possible. The assessment standard asks what information a rule receives, what its verdict claims, and which assumptions support it. The account relates the constraint to Dretske’s treatment of information and certainty and examines the 2008 construction of a rogue certificate authority certificate using an MD5 collision. The signed portions of two different certificates shared a digest, allowing a genuine signature to verify against both. The example concerns the information used in the signature-verification step. The researcher demonstration does not establish how often such collisions were exploited. Part of the Equivalence and Interchangeability series within On Human Modeling, the paper offers questions for assessing what verification can establish about a record’s history. Its framework can inform human–AI research on reconstructions from incomplete records, the assumptions supporting probabilistic judgments, and the certainty users attribute to verification results.