AI evidence extraction

Modelo conceptual y herramienta abierta para la declaración estandarizada del uso de IA generativa en la producción científica

German Díaz Hernández · 2026

AI-generated evidence extraction, verified across multiple analytical personas. Not a substitute for the peer-reviewed original.

9/10

Relevance

2/4

Quality (LMQS)

Evidence

Citations

0.00

FWCI

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.62059/latarxiv.preprints.634

Methodology & findings

Study design

Mixed methodology: (1) Exploratory (non-systematic) literature review of normative and ethical requirements (2023-2025) across PubMed, Scopus, arXiv, and institutional sources (COPE, ICMJE, UNESCO); (2) Technical platform analysis of usoeticoia.org functional characteristics and governance; (3) Mapping of integration points with academic information systems (OJS, DSpace, institutional repositories); (4) Deployment and preliminary adoption signal collection during first operational week (December 2-10, 2025)..

Primary method

Design science research (DSR) with elements of participatory/human-centered design. The artifact operationalizes a conceptual model developed through normative analysis (literature review of COPE, ICMJE, UNESCO guidance) and technical requirements mapping. No formal user co-design process is reported; design principles are driven by identified normative gaps and technical interoperability requirements.

Main result

During the first week of operation (December 2025), the platform generated 8 declarations and collected 16 public signatories from 16 countries and institutions. The paper states: "En su primera semana de operación (diciembre 2025), la plataforma registra señales preliminares de adopción: 8 declaraciones generadas y 16 firmantes del compromiso ético provenientes de 16 países e instituciones distintas." The study proposes a conceptual model structured around three components: a taxonomy of uses distinguishing between assistance, co-creation, and substantive transformation; an operational human review scale with six levels (0-5); and a dual output structure with cryptographic verification.

Research paradigm

Design science; pragmatist (human-centered design for addressing normative gap in AI disclosure practices)

Author conclusions

The authors conclude that the conceptual model operationalized by usoeticoia.org addresses the 'transparency paradox' documented by Schilke and Reimann (2025) through design: "La hipótesis que sustenta usoeticoia.org es que estos problemas son remediables mediante diseño." They propose that structured, differentiated taxonomies combined with visible human oversight and standardized formats can normalize disclosure practices. The authors state the tool is "diseñada para beneficiar a: personas autoras (declaración proactiva), editores (verificación de cumplimiento), comités editoriales (estándar para políticas), instituciones educativas (herramienta pedagógica), estudiantes de posgrado (documentación en tesis), repositorios (metadatos estandarizados) y agencias de financiamiento (seguimiento de uso de IA)."

Risk of bias

Self-selected sample (users who discovered the platform); Regional bias toward Iberoamerica due to initial dissemination; Very narrow temporal window (7 days); No control group or baseline for comparison; Cannot infer causal impact or population adoption rates; Selection bias (self-selected users discovering the platform); geographic bias toward Iberoamerica due to initial dissemination strategy; temporal limitation (7-day observation window); lack of causal inference capability.; Selection bias: Self-selected sample of users who discovered the platform; Geographic bias: Regional bias toward Iberoamerica due to initial dissemination strategy; Temporal bias: Extremely short observation window (7 days) prevents population-level inference; Verification bias: Tool depends on user honesty; cannot directly verify declared AI use matches actual use; Language/access bias: Platform operates in Spanish, English, Portuguese, Italian; non-representative of global academic community; Reporting bias: Only users who chose to register declarations are counted

Limitations

The authors explicitly state: "Limitaciones: muestra auto-seleccionada, sesgo regional hacia Iberoamérica, ventana temporal reducida (7 días)
No se pueden inferir tasas de adopción poblacional ni impacto causal." They further note that "Estos datos deben interpretarse como señales preliminares de interés, no como evidencia de adopción masiva." Additional limitations include: "Esta revisión no constituye una revisión sistemática con protocolo PRISMA" and the inability to directly verify whether "la información declarada refleja fielmente el uso real."

Open questions raised

Empirical validation of the hypothesis that structured taxonomies and human oversight scales can mitigate the transparency paradox documented by Schilke and Reimann (2025)
Consensus international standard schema for AI-use metadata in academic publishing
Migration to JSON canonical serialization (RFC 8785 / JCS) for hash verification interoperability
Full legal compliance audit across multiple jurisdictions
Usability and adoption studies among authors, editors, and institutional repositories
Integration plugins for OJS 3.x and DSpace (planned Q2 2026 and beyond)

Data: Research data are included in the manuscript itself. The paper states: "Los datos de investigación están incluidos en el propio manuscrito."Code: https://github.com/escire/usoeticoia-declarador; https://github.com/escire/usoeticoia-declarador/blob/v1.1.0/core/schema.json; https://github.com/escire/usoeticoia-declarador (main repository); https://github.com/escire/usoeticoia-declarador/blob/v1.1.0/core/schema.json (JSON schema source of truth)Extracted from: pdfAgreement 54%

Explore related topics

Institutional Policy AI Governance Design Science