Italian Elections and Political Communication
From Hybrid TV–Twitter Publics to Platform-Governed Visibility
The arc filed under this topic traces a decade-long empirical investigation of how Italian elections are mediated, manipulated, and measured on social platforms. The earliest entry, Iannelli2015-e0818c3e, examines the 2013 campaign as a hybrid phenomenon in which Twitter operated as a second screen to political talk shows, and finds that despite millions of tweets, participation was narrow, audience-driven, and largely subordinate to TV logic. That paper sets the analytic baseline against which the subsequent MINE work develops: a polarized-pluralist media system in which networked publics amplify rather than displace legacy broadcast agendas, and in which political actors fail to meaningfully engage their audiences.
Partisanship, Insularity, and the Populist Attention Economy
The 2018 cycle marks a shift from hybrid-media descriptions to structural analyses of partisan attention. Giglietto2019-882f1900 combines Twitter-derived insularity scores with Facebook engagement to show that populist-aligned sources (especially M5S) are markedly more insular, and that insularity correlates with a sharing-over-commenting pattern indicative of echo-chamber amplification rather than contestation. M5S emerges as uniquely adept at the social-media attention economy — a finding that prefigures later interest in coordinated amplification. Insularity and partisan asymmetry, here treated as emergent properties of audience behavior, are reframed in subsequent work as the targets of deliberate coordinated action.
From Content to Coordinated Behavior
Giglietto2020-9d8acdd7 formalizes that turn, proposing Coordinated Link Sharing Behavior (CLSB) as an ecological, action-based alternative to content- and actor-based disinformation detection. Across 2018 and 2019, coordinated networks were shown to disproportionately amplify problematic domains, with deceptively non-political networks exhibiting narrower domain concentration than openly political ones. Giglietto2023-fa71a001 extends this paradigm methodologically: static lists of bad actors decay quickly, so the 2022 election workflow iteratively surfaces new coordinated accounts in near-real-time. The case studies — an M5S hyperpartisan echo network, a religious-clickbait click-economy operation, and a Church of Almighty God proselytism cluster — demonstrate that the same behavioral signature surfaces ideologically, economically, and religiously motivated operations alike.
Methodological Consolidation: LLMs in the Loop
As the corpus of Italian political content grew across cycles, the programme increasingly turned to LLM-based pipelines for processing. Giglietto2024-cbeb3f70 compares embedding models for unsupervised clustering of 2018 and 2022 political news, finding that OpenAI’s text-embedding-3-large consistently outperforms the Italian-specific UmBERTo, and that clustering quality declined between the two election years — a substantive signal hiding in a methodological paper. Marino2024-2fbc690f generalizes this into a reflective account of LLMs-in-the-loop validation, arguing that classification, embedding, and labeling each require their own validation step, and that expert (not crowd) annotators are necessary because LLMs already outperform low-skilled coders on nuanced Italian political tasks. Together these papers operationalize the analytical infrastructure underlying the broader empirical programme.
Platform Governance Enters the Frame
The most recent entries — the twinned working papers Giglietto2025-1765bb4f and Giglietto2025-1e9a0917 — pivot the topic from studying political actors on platforms to studying the platforms themselves as political actors. Using DSA-enabled Meta Content Library data and structural breakpoint detection, they show that Meta’s political content reduction policy was effectively implemented in Italy roughly ten months before its announced global rollout, cutting re-elected MPs’ average per-post reach by 72%. The January 2025 reversal restored only ~65% of baseline. Crucially, the policy’s effects were asymmetric: extremist accounts compensated through sharply increased posting frequency and ultimately gained aggregate weekly reach, while mainstream politicians compounded their per-post losses by posting less. This finding closes a circle with the earlier work — the very hyperpartisan ecologies documented in Giglietto2020-9d8acdd7 and Giglietto2023-fa71a001 turn out to be structurally advantaged by the platform’s own depoliticization policy.
The Arc
Read together, the eight papers trace a clear trajectory: from describing hybrid TV–Twitter publics (2013), to mapping partisan insularity (2018), to detecting coordinated amplification (2018–2022), to industrializing LLM-based analysis of the resulting corpora, to scrutinizing the platform-level algorithmic interventions that reshape who is visible at all. The unifying claim across the programme is that Italian electoral communication cannot be understood at any single layer — content, actor, audience, or algorithm — but only through behavioral and infrastructural analyses that move with the platforms themselves.