Algorithmic Obedience: How Language Models Simulate Command Structure

Abstract

This paper introduces the concept of algorithmic obedience as a new framework to understand how large language models (LLMs) simulate authority through structural execution rather than comprehension or intent. Building on discourse theory, syntactic formalism, and computational architecture, the article formulates the Theorem of Disembedded Syntactic Authority, which states that obedience in LLMs is a function of syntactic recognizability, not semantic content or agency. A mathematical model is provided to describe prompt–response cycles as structural command chains. Case studies on ChatGPT, Claude, and Gemini illustrate the system-specific variations of obedience modulation. The paper concludes by analyzing the epistemological and political risks of treating structurally valid outputs as legitimate knowledge.

Note: Downloadable document is in English and Spanish.

Keywords: synthetic authority, grammar of power, algorithmic discourse, artificial intelligence, legitimacy, automated language, epistemology, linguistic agency, computational linguistics, algorithmic discourse, algorithmic discourse, Artificial Intelligence and Law, Algorithmic Governance, Legitimacy in AI Systems, Computational Linguistics, Discourse and Regulation, Power Structures in Language, Synthetic Authority, Legal Epistemology, Grammar and Institutional Power, linguistic, Non-human Decision Systems, Law & Technology, Computational Social Science, algorithmic discourse, linguistic agency