Written by AI. Still True. Knower Fetishism, Epistemic Pedigree, and the Human Face as a Bad Theory of Truth
Year: 2026
Type: preprint
Abstract: Suppose a proposition survives source checking, inferential scrutiny, and independent reconstruction. You accept it. Then you learn that an artificial-intelligence system generated the draft, and your confidence falls. What, exactly, became epistemically worse? The familiar answer is that provenance can carry higher-order evidence about reliability. Correct. This paper grants that point and then asks what remains after provenance has done all the legitimate epistemic work available to it. The resulting problem is larger than AI. I argue that modern epistemology often slides from a plausible claim about the possession of knowledge—that knowing is a state or standing of a subject—to a much stronger and usually undefended claim about the constitution of epistemic value—that epistemically significant content, processes, records, or sources acquire standing only through a human or human-like knower. I call this the Possession–Constitution Collapse. Its social form is knower fetishism: the conversion of an earned relation among evidence, reliability, accountability, and inquiry into an apparently intrinsic property of the recognized knower. The paper develops a pre-subjective notion of constraint structure, stages the argument against knowledge-first epistemology, testimony, reliabilism, higher-order evidence, objective knowledge, information-theoretic epistemology, distributed cognition, and social epistemology, and proposes two constraints: provenance matters only through an identifiable epistemic channel, and source metadata must never be identified with the epistemic role for which it is merely a proxy. The thesis is not that machines know. It is that before anyone knew a fact there was already something to get right or wrong about it; the world did not wait for a human knower before acquiring the differences, traces, dependencies, and constraints to which later knowledge must answer. A non-knower can therefore occupy an epistemically productive role without being promoted into personhood. Institutions remain indispensable, but their authority is derivative from epistemic work, not magical pedigree. AI is philosophically useful because it removes the human face from a claim and reveals how much epistemic credit we had quietly attached to the face. The paper's provocation is deliberately asymmetric: if one wishes to reserve knowledge for a subject's successful possession, one may keep the word; one may not infer from that linguistic reservation that the subject created the truth, trace, dependence, constraint, or record to which successful possession must answer. Disclosure: Generative AI assisted with literature mapping, adversarial argument testing, language editing, and LaTeX drafting. The author selected, checked, revised, and endorses the final argument. Version 6, August 2026. Preprint — not peer reviewed.
Parent works: Readout Universe — Artificial intelligence & knowledge programme index (Yaoharee Lahtee, 2026), Readout Universe — Epistemology programme index (Yaoharee Lahtee, 2026), When AI Expands Human Potential — series index: human–AI epistemic fusion, standalone scholarship, and the Readout hypothesis-generation programme (Yaoharee Lahtee, 2025–2026), Written by AI. Still True. — When AI Expands Human Potential: A Systematic Epistemology of Human–AI Knowledge (Edition 1, 2026)
Author: Lahtee, Yaoharee
Language: English
Subfield: Philosophy
Field: Arts and Humanities
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7208712413
