Generative AI is shifting epistemic authority from people and institutions to opaque algorithmic infrastructures, where probabilistic fluency often substitutes for accountable expertise. That shift both privatizes aspects of the linguistic commons into corporate architectures and alters how users direct trust and desire toward non‑subjective systems.

Main Finding

Generative AI (LLMs) does not literally become a new epistemic subject, but it functionally reconfigures where and how symbolic authority is located. Through probabilistic fluency, responsive interfaces, and institutional embedding, LLMs are being placed in the structural role analogous to Lacan’s “Big Other”: a presumed locus of knowledge and guarantee that is infrastructural, opaque, and procedural rather than embodied, accountable expertise. This shift encloses linguistic commons within proprietary architectures, redistributes epistemic trust, shapes subjectivity (transference, compulsive use), and makes hallucinations the visible return of underlying structural lack.

Key Points

• Functional—not ontological—displacement of authority:
  • LLMs substitute for sites of verification and synthesis when users and institutions routinize reliance on their outputs. They are treated as a source of answers without actually possessing judgment, responsibility, or intentionality.
• Three mechanisms that enable this positional shift:
  • Probabilistic fluency: fluent, context-sensitive outputs create an impression of comprehensive knowledge.
  • Interface responsiveness: conversational UX encourages address and projection of authority.
  • Institutional embedding: routine incorporation into pedagogy, research, journalism, administration routinizes AI outputs.
• From institutional to infrastructural authority:
  • Traditional legitimating structures (visible institutions, embodied expertise) are being supplanted by distributed, opaque computational infrastructures whose legitimacy rests on indispensability and fluency rather than procedural accountability.
• Political economy and enclosure:
  • LLMs centralize linguistic capital into proprietary platforms (data enclosure), extending platform/surveillance capitalism into discursive production and producing new rents tied to model access and control.
• Subjectivity and affective dynamics:
  • Generative interfaces engender transference-like relations and compulsive engagement (a human “jouissance” bound to algorithmic repetition).
• Hallucination as structural signal:
  • Model hallucinations are interpreted psychoanalytically as the return of structural lack—they expose the impossibility of complete epistemic closure and serve as ruptures in algorithmic authority.
• Normative observation:
  • The problem is not machine consciousness but the redistribution of epistemic authority and its governance, transparency, and political-economic consequences.

Data & Methods

• Methodological approach: conceptual/theoretical synthesis and critical interpretation.
  • Primary framing: Lacanian psychoanalysis (Big Other, lack, transference) combined with media theory, political economy, and philosophy of technology.
  • Evidence base: literature review across AI studies, media theory, sociology, political economy, and examples from journalism, education, clinical settings, and platform studies (cited empirical studies used illustratively but no original quantitative data).
  • Analytical moves: structural-functional argument (how practices of reliance reassign authority), phenomenological description of user–AI interaction, political-economic analysis of enclosure and platformization.
• No original empirical dataset or econometric analysis; claims are argued via interdisciplinary theoretical reasoning and by synthesizing contemporary empirical literature.

Implications for AI Economics

• New source of economic rents
  • Procedural/infrastructural authority becomes monetizable: firms owning high-quality training corpora, APIs, and UX can capture value via subscription, platform lock-in, and network effects.
  • “Fluency” is an economic asset: perceived reliability drives demand, even when epistemic grounding is weak—creating rents that may be decoupled from factual correctness.
• Market structure and concentration
  • Data enclosure and proprietary model stacks increase entry barriers and concentration in a small number of platform owners, reinforcing winner-take-most dynamics and platform control over discursive markets.
• Labor and epistemic work
  • Reconfiguration of epistemic labor: some tasks (summarization, drafting, backgrounding) may be deskilled or reallocated, while verification, curation, auditing, and oversight gain value.
  • New labor demands for model auditing, prompt engineering, fact-checking, and institutional governance roles.
• Externalities and trust costs
  • Hallucinations and opaque provenance create negative externalities (misinformation, liability, reputational costs) that are not internalized by platforms unless regulated or priced.
  • Institutions relying on LLMs may experience hidden costs from errors, requiring investment in verification and insurance markets.
• Regulatory and policy levers with economic consequences
  • Transparency requirements (model cards, provenance, audit trails) can reduce asymmetries and alter competitive advantages.
  • Data-rights, compensation for content creators, and limits on data enclosure would affect platform inputs and model valuation.
  • Standards for disclosure/citation of AI-generated content could change the unit economics of content production and distribution.
• Public goods and alternative provisioning
  • Open models, commons-based datasets, and cooperative infrastructures could counteract enclosure, alter pricing dynamics, and re-distribute rents.
• Research and measurement priorities for AI economics
  • Quantify the economic value of "probabilistic fluency" and the price of perceived authority versus factual correctness.
  • Empirically measure the cost of hallucinations to firms/institutions (reputational, legal, operational).
  • Model how data enclosure affects competition and welfare in markets for information and knowledge services.
  • Study labor-market transitions for epistemic professions and the demand for verification/audit services.

Suggested policy/economic responses (brief) - Promote transparency, provenance, and auditability to better align price signals with factual reliability. - Consider antitrust scrutiny of data and model-stack vertical integration. - Create mechanisms for compensation or opt-outs for creators whose work fuels proprietary models. - Fund public or cooperative models to supply a non-enclosed baseline of discursive infrastructure.

Reference - Hamamra, B., & Uebel, M. (2026). Generative AI and the Reconfiguration of Symbolic Authority: Implications of a Psychoanalytic View. Human Arenas. https://doi.org/10.1007/s42087-026-00687-y