Meta-Identities and Political Extremism in Digital Networks: Algorithmic Classification, Radicalization and Democratic Disputes

Abstract

This article examines how the literature documents classificatory processes produced by digital platforms and artificial intelligence systems about subjects, content and collectives involved in radicalized political disputes. Meta-identity is understood as a sociotechnical classificatory construct — inferred, aggregated, opaque and operational — produced about entities circulating within digital infrastructures, rather than by them, and mobilized as a reference for recommendation, moderation, visibility, segmentation and suspicion. The central question is how the literature documents classifications that may favor, limit or reorganize the circulation of extremist and anti-democratic repertoires. The procedure is a structured evidence synthesis: 813 records retrieved from open bibliographic sources between 2016 and 2026, consolidated into 749 studies, of which 127 had at least one full text retrieved; in 118 of them, rule-based, auditable lexical localization identified at least one candidate. The search did not look for the concept of meta-identity, but for the phenomena it analyses and describes, using the vocabulary of the original literature. Preliminary results indicate that the literature readily documents the inference and stabilization of classifications, less frequently their attribution to determinate bearers, and almost never exteriority — the condition whereby the classified party lacks substantive control over the production and revision of the predicate. Adversarial activation of input dynamics by organized political actors appears in a small number of studies. We conclude that democratic dispute involves not only the regulation of extremist messages, but the contestation of the classificatory regimes that define what will be seen, recommended, moderated or rendered suspect.