Generative AI as External Operational Mediation: Open Research Materials, Supplementary Methodological Guide, Figures, and Coded Corpus Dataset (v15)

This Zenodo record provides the open-access research materials, supplementary methodological documentation, figure package, and coded corpus dataset associated with the manuscript:

“Generative AI as External Operational Mediation: A Conceptual Integrative Review of Cognitive Externalization and Human-AI Division of Labor”

Author: Roman R. Sidorchuk  
Affiliations: Plekhanov Russian University of Economics; RUDN University  
ORCID: 0000-0002-4033-2937  
Correspondence: Sidorchuk.RR@rea.ru  
Version: v15  
DOI: 10.5281/zenodo.20026337  
Publication date of this record: 2026-05-04  
License: Creative Commons Attribution 4.0 International (CC BY 4.0)  
Copyright: © 2026 Roman R. Sidorchuk

This record is an open research materials package prepared to support transparency, reproducibility, independent inspection, and citation of the conceptual integrative review. The package is not a substitute for the journal submission itself and should be understood as an open-access research object containing the author manuscript version, supplementary methodological documentation, coded literature corpus, figures, metadata files, and reproducibility-oriented materials.

The manuscript conceptualizes generative artificial intelligence (GenAI) as a form of external operational mediation within human-AI cognitive systems. The central argument is that GenAI should not be reduced to text generation, productivity enhancement, or digital writing. Instead, the review frames GenAI as an external computational environment that can participate in intermediate cognitive operations: drafting, structuring, comparison, interpretation, alternative generation, option narrowing, decision preparation, and, at the frontier, metacognitive support.

The article contrasts writing as the historically established first wave of cognitive externalization with GenAI-enabled operational delegation as a second-wave form of externalization. Writing is treated as an external medium that fixes, stabilizes, stores, and enables repeated access to meaning. GenAI, by contrast, is interpreted as an external operational mediator capable of transforming incomplete task formulations into drafts, decompositions, criteria, comparisons, recommendations, and action-oriented suggestions. This distinction is used to clarify how human-AI systems redistribute cognitive labor without eliminating human responsibility for final judgment, validation, and use.

The conceptual framework is organized around the Cognitive Externalization Scale (CES), an author-developed analytical coding scheme used to classify levels of functional redistribution between the human subject and the external cognitive circuit. CES is not presented as a psychometric instrument or validated psychological scale. It is a functional coding framework designed for literature synthesis. The four CES levels distinguish: external fixation and repeated access to meaning; external participation in preparatory cognitive operations; external narrowing or coordination of alternatives; and external involvement in metacognitive regulation, memory governance, reflective control, or confidence calibration.

The article also formulates the Third Signaling System (TSS) as a boundary hypothesis for AI-mediated metacognitive regulation. TSS is not proposed as a neurophysiological claim and is not equated with consciousness or autonomous cognition. It is defined operationally as a possible external metacognitive circuit: a computational environment that may participate in memory management, uncertainty monitoring, confidence calibration, and the organization of symbolic operations performed with a human subject. In the terminology of the manuscript, TSS corresponds to the CES-4 boundary, where an external system begins to affect regulatory loops that were previously treated primarily as human metacognitive functions.

The review is positioned as a conceptual integrative review and theory synthesis. It does not claim to be a registered systematic review, a full bibliometric review, or a scoping review. PRISMA-inspired selection-flow diagrams are used as transparency aids to document identification, screening, eligibility, and inclusion logic, without implying a registered systematic-review protocol. The synthesis follows a concept-centric logic and organizes the literature around functional mechanisms rather than isolated study summaries.

The research materials include four coded literature corpora. Corpus 1 addresses writing, literacy, external memory, and the cognitive effects of writing. Corpus 2 addresses GenAI, cognitive offloading, decision support, structured prompting, executive functions, productivity, work redesign, and AI writing. Corpus 3 addresses memory, forgetting, metacognition, reflective control, memory governance, and AI-assisted memory coordination. Corpus 4 addresses AI-mediated choice, consumer decision-making, recommendations, personalization, conversational interfaces, trust, and the economics of AI. Together, these corpora support a staged interpretation of cognitive externalization from external memory to operational mediation and, at the frontier, to possible external metacognitive regulation.

The coded corpus dataset contains 90 analytical records across the four corpora. Each record includes corpus assignment, bibliographic and metadata fields, thematic coding, CES level, functional interpretation, and evidence-status notes. The dataset is provided in both XLSX and CSV formats. The XLSX file is the preferred machine-readable version for inspection and reuse; the CSV file is provided as a plain-text fallback format.

The supplementary methodological guide documents the search logic, corpus-level coding rules, CES codebook, coding prompts, stability indicators, and synchronization notes for the v15 manuscript. It clarifies that the coded corpus is a secondary literature coding dataset and that CES levels are author-developed functional labels rather than psychometric measurements. It also documents the role of AI-assisted comparison in the intra-rater coding stability check. The generative model was not treated as an independent reviewer, judge, or co-coder. It was used only as a technical instrument to compare two temporally separated authorial coding passes, identify discrepancies, and flag boundary cases for authorial review. Final coding decisions, CES assignments, source inclusion decisions, and theoretical interpretations remained authorial.

The package also includes high-resolution figures used in the manuscript, a reproducibility-oriented methodological package, README files, citation metadata, Zenodo metadata, Figshare-oriented metadata, license information, checksums, and a separate concept note summarizing the article’s conceptual framework, propositions, conclusions, and copyright statement.

The main conceptual conclusions documented by the package are as follows. First, writing provides the historically established baseline for cognitive externalization because it externalizes memory, fixation, and repeated access to meaning. Second, GenAI introduces a qualitatively different form of externalization by shifting intermediate cognitive operations into an external computational circuit. Third, this shift should be understood as a redistribution of the human-AI division of labor rather than as a replacement of the human subject. Fourth, AI-mediated choice systems, recommendation interfaces, conversational agents, and decision-support environments demonstrate the applied market-facing consequences of operational mediation. Fifth, metacognitive regulation remains a frontier rather than a fully established dominant pattern of current GenAI use. Sixth, the Third Signaling System hypothesis is best understood as a boundary hypothesis for future research on external metacognitive regulation, not as an ontological claim about machine consciousness.

These materials may be used for scholarly inspection, replication of the coding logic, teaching, methodological review, conceptual comparison, and future research on generative AI, cognitive externalization, cognitive offloading, human-AI interaction, distributed cognition, metacognitive regulation, AI-mediated decision support, and market-facing choice architectures.

Recommended citation:

Sidorchuk, R. R. (2026). Generative AI as External Operational Mediation: Open Research Materials, Supplementary Methodological Guide, Figures, and Coded Corpus Dataset (v15). Zenodo. https://doi.org/10.5281/zenodo.20026337