The Epistemic Black Hole: Why the Deletion of Ballot Images Violates the Foundational Chain of Custody in Regulated Engineering Systems
Year: 2026
Type: preprint
Abstract: In safety-critical, regulated engineering disciplines, the deterministic transformation of an analog physical artifact into an actionable tabular classification relies on unbroken data provenance. When an optical sensor ingests a physical object to generate an automated determination, foundational quality control (QC) frameworks—such as ISO/IEC 17025, GAMP 5, and AS9100—dictate that every intermediate representation in the computational chain must be immutably preserved. Purging intermediate image rasters severs the causal link between sensor input and algorithmic output, rendering downstream tabular metrics mathematically unverifiable. This paper demonstrates that the deletion or discretionary retention of ballot images violates the foundational chain-of-custody principles established across mission-critical systems engineering. We examine three benchmark industrial domains—aerospace computed radiography and optical inspection, pharmaceutical fill-finish parenteral inspection, and forensic 3D ballistic profilometry—detailing their end-to-end chain-of-custody protocols, the specific federal statutes and international quality standards they satisfy, and the catastrophic legal and operational consequences that occur when raw sensor records are discarded. We contrast these rigorous architectures with optical-scan election tabulation, where disparate state-level practices have historically treated digital ballot images as optional, ephemeral cache. Finally, we analyze the systemic auditability risks of image destruction under federal preservation law (52 U.S.C. § 20701), demonstrate how overlapping jurisdictional evasions create a closed loop of administrative unaccountability, and establish a mandatory engineering specification that eliminates administrative retention toggles in favor of cryptographically bound, dual-layer verification.
Author: Ayyadurai, V.A. Shiva
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7225310106