The Production and Control of Representations in Artificial Intelligence — Research II: A Conceptual Architecture, Authority Structure, Representation Lifecycle, and Research Protocol
Research II examines the structural relations through which representations are constituted before they reach a recipient, treating representation itself as the central analytical object rather than treating the model, the platform, or the user in isolation. The program's central question is: who produces the representations a recipient encounters through AI systems, who controls their formation, availability, and appearance, and who consumes them? This question is structural: it does not presuppose that production, control, and consumption are each located in a single entity. The program specifies four foundational documents: 1. A Conceptual Architecture (adopted, version 0.2) distinguishing five functional units — Producer, Model/Generative System, Platform/Controller, Distribution Interface, and Recipient/Consumer — defined by function rather than by institutional identity. 2. An Authority Structure (version 0.1, pending adoption) distinguishing five authority types — Production, Transformation, Selection, Distribution, and Interpretive Authority — and establishing that authority is analytically distinct from function. 3. A Representation Lifecycle (version 0.1, pending adoption) tracing six structural stages through which a representation may pass, specifying a canonical path and five structurally distinct alternative paths, together with the conditions under which the lifecycle may be interrupted at each stage. 4. A Research Questions document (version 0.1, pending adoption) formulating twenty research questions across four families — conceptual, authority, lifecycle, and boundary questions — including an explicit boundary question concerning the relation between Research II and a separate, methodologically independent research program (Research I) on epistemic discipline in human-AI interaction. No empirical work has been performed under this protocol. The literature review, the methodological scope, and the selection of systems for analysis remain to be determined following adoption of the four foundational documents. The program's commitments are therefore reported as specified, not as established: this repository presents an architecture and a research protocol, not findings. Version 0.1.0 — Initial Program Manuscript — Architecture Specification Stage. November 2026.
Authors
- Samir Baladi (ORCID: https://orcid.org/0009-0003-8903-0029)
Publication Details
- Journal
- Figshare
- Published
- 2026-11-04
- DOI
- https://doi.org/10.6084/m9.figshare.34908597
- Primary Topic
- Ethics and Social Impacts of AI
- Type
- preprint