The Invisible Retention Rate (IRR): An Unvalidated Candidate Formalization for Examining Information Retention Across Representation Processes

Version 2 replaces the previous version of this document. The earlier version described the Invisible Retention Rate (IRR) as a “formal metric” that “measures” the preservation of invisible informational elements and associated higher IRR values with architectures capable of preserving epistemic plurality. Those formulations exceeded what the conceptual and empirical basis of the proposal could support. Version 2 presents the IRR more narrowly as an unvalidated candidate formalization. It is proposed for examining whether information elements defined in advance remain traceable after a representation or processing step. Within a specified coding scheme and matching rule, the candidate formula expresses the proportion of designated input elements that can still be identified in an output. The IRR has not been validated as a measurement instrument, diagnostic tool, or indicator of ethical quality. It does not directly measure invisibility, empathy, fairness, epistemic value, or ethical quality. Reliability, construct validity, matching accuracy, thresholds, and practical utility have not been established. Readers should use and cite Version 2 rather than the previous version.