Jonathan d'Aquitaine · Zenodo (CERN European Organization for Nuclear Research) 2026 · 2026
DOI: 10.5281/zenodo.22851978
Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).
Large language models and AI-mediated search systems have become the primary intermediaries for institutional information across educational, legal, civic, and professional life. Current models of AI ethics focus primarily on the accuracy, bias, and fairness of AI systems. This focus is essential but fails to recognize a structural failure that precedes accuracy: in an information context, where the predominant risk to epistemic integrity is not misinformation or error but the fabrication of institutional authority, AI systems are designed to produce coherent, confident, and complete responses. There is a structural match between the optimization process and the associated threat. This paper introduces two theoretical constructs to name this problem. Epistemic totalization is the first type, in which the communicative functions of inquiry, assertion, interpretation, and authority are lost in a total function that lacks concession, appeal, or authorship. The second, the distinction between fabricated and genuine competence, redirects the difference between real and performed expertise from an epistemic to a behavioral and psychological dimension, genuine competence that shows convergence when challenged, and fabricated competence that shows escalation. The paper also builds on theories of embodiment in the performing arts to suggest that the AI confirmation systems are an ideal, non-threatening, confirming audience for a particular, less-theorized psychological state that differs from what Carl Rogers would call a stable self, in which professional performance becomes a part of identity instead of a role adopted over a stable self. The paper argues that there is one enduring boundary that current AI systems cannot breach: the governmental chain of custody, which includes court dockets, licensing registries, federal identifier databases, sworn filings, and other processes that precede individual claims and require accountable human beings to record facts rather than merely confirm presentations at each step. The case study is presented as an unadjudicated, anonymous example from a public record; it appears as evidence in the process of collapsing, link by link, as the governmental terminus approaches, but it is never considered a fact but rather evidence that is as much a tool for self-referential credential architecture as a means of its destruction. The paper argues that ethical communication frameworks should also change with respect to what an AI system is structurally unable to do: defer to unreachable authority, note records it cannot access, and refuse unverified confirmation. This argument does not end the identification of pluralism in credentialing and education; it makes it possible. Inclusion is only a pluralistic approach with chain-of-custody integrity. Standards cannot replace pluralism: the process of learning can be flexible, but the content of learning cannot be negotiated. Keywords: AI ethics, epistemic totalization, credential fabrication, chain of custody, impostor phenomenon, performance theory, identity, algorithmic confirmation, pluralistic education, epistemic authority
No comments yet — start the discussion below.