From Kant to "Agents of Chaos" Why the Central Problem of AI Is Not Intelligence but Epistemic Authority

Read full text
Authors: Markus Dören
Cited by: 2

Abstract

Contemporary debates surrounding artificial intelligence increasingly focus on capability, automation, alignment, and machine consciousness. This paper argues that the deeper societal challenge of modern probabilistic semantic systems may lie elsewhere: in the gradual transfer of epistemic and operational authority from humans and institutions toward semantically persuasive but fundamentally non-grounded systems. The paper develops a conceptual and governance-oriented analysis connecting classical epistemology, philosophy of language, cognitive psychology, semiotics, systems theory, and contemporary AI governance discourse. By tracing a conceptual line from Immanuel Kant’s theory of judgment, through Ludwig Wittgenstein’s language games, Umberto Eco’s semiotics, Joseph Weizenbaum’s ELIZA effect, Daniel Kahneman’s dual-process cognition, and Peter Kruse’s network dynamics, toward modern agentic AI architectures and multi-agent instability phenomena such as “Agents of Chaos,” the paper argues that the defining challenge of the AI era is not machine intelligence itself, but the institutional consequences of semantic persuasion. The paper introduces the concepts of Cognitive Surrender, Authority Drift, Epistemic Dependency, Semantic Lock-In, and Structural Fragility as governance-oriented analytical constructs describing emerging risks within AI-mediated institutional environments. It further proposes the principle of Semantic Authority Separation (SAS), arguing that semantic generation and operational authority should remain structurally separable in stable human–AI decision architectures. Rather than focusing primarily on model capability or machine ontology, the paper reframes AI governance as a problem of preserving human and institutional judgment under conditions of deep probabilistic semantic mediation. The central governance challenge of the AI era may therefore not be artificial intelligence itself, but epistemic authority.

Fields of Study & Subjects

Embodied and Extended Cognition Ethics and Social Impacts of AI Cybernetics and Technology in Society

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
25 May 2026
Language
EN
Document type
Preprint