Algorithmic Classifications and Democratic Discourse: Restricted Research Dataset

Résumé fourni par la source

This restricted dataset contains the empirical, methodological, and technical materials associated with the study “Algorithmic Classifications and Democratic Discourse: Meta-Identity as an Analytical Model for the Study of Polarisation on Digital Platforms.” The project applies a mixed-methods grey-box strategy to publicly observable testimonies, platform-response claims, moderation disputes, and documentary evidence concerning political content. Its geographical scope covers Brazil, Portugal, Spain, Argentina, and Italy. The design includes the initially selected political actors, their parties and political ecosystems, and symmetrically defined national opposition poles. Searches were conducted in Portuguese, Spanish, Italian, and English. The deposit contains an integrated multiplatform dataset, a Reddit-specific collection layer, a separate media and litigation dataset, machine-readable CSV and JSON files, query dictionaries, political scopes, validation decisions, collection and assembly code, and technical provenance. The consolidated provenance documents 12,248 technical operations. The integrated record-level dataset contains 94 screened records: 1 validated Class A record, 3 Class B records, and 90 Class C controls or exclusions. The media and litigation layer contains 19 documentary records. The dataset documents observable outputs, retrieval conditions, exclusions, and structural limitations of platform access. It does not provide access to proprietary classifiers, does not establish whether individual decisions were human, automated, or hybrid, and does not support prevalence estimates or causal claims of partisan discrimination. Files are restricted because they contain public-source URLs, text fragments, search snippets, contextual information, and technical traces that may facilitate source identification or the reconstruction of deleted or unavailable content. Access may be granted for non-commercial academic or public-interest research after review by the dataset creators through the project account: metaidentidade@gmail.com. Approved access is purpose-specific, personal, non-transferable, revocable, and subject to secure handling, non-reidentification, and non-redistribution conditions. Redistribution, contacting source users, reconstruction or republication of deleted content, political targeting, commercial surveillance, and model training are prohibited.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.