Retention Limits on Later Discrimination: How What a Process Keeps Bounds What Can Later Be Told Apart
Year: 2026
Type: article
Abstract: Later analysis can be careful and still be unable to make a distinction that an earlier process removed. This paper sets out Retention Limits on Later Discrimination, a structural constraint on what subsequent operations can establish from a retained record. When a retention rule maps relevant source differences to the same output, further processing of that output alone cannot recover the merged distinction. The paper distinguishes genuinely lost distinctions from differences that have merely been degraded and remain partly recoverable. It examines how independent information, including an original record, another identifier, or relevant contextual knowledge, can supply what the retained value no longer contains. A retention rule is assessed by whether it preserves the distinctions required by later uses and leaves another route available when an omitted distinction becomes important. The discussion relates this constraint to research on data provenance and examines the automatic conversion of gene symbols by spreadsheet software. The case illustrates loss within converted cell values while recognizing that other fields or context may still allow the original gene to be identified. Part of the Relational Conditions and Dynamics series within On Human Modeling, the paper identifies questions for examining what a process keeps, what later work requires, who sets the retention rule, and who bears the consequences of information loss.
Author: Cary, Alexandra
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7225637422
FWCI:
Cites:
Cited by:
Related to: