Uni Urbino Team Members at Conferences in Brazil: Overview of Activities
In October 2025, we (team members of the University of Urbino) presented a variety of research outputs related to vera.ai across two major conferences in Brazil. Events brought together scholars from all over the world. A shared aim: examine how disinformation spreads, evolves, and shapes our digital landscapes. Here's a recap and some information about how we contributed to these events.
International Seminar on Disinformation, Social Media, and Climate (Porto Alegre, October 9–10)
Our first stop was the Federal University of Rio Grande do Sul in Porto Alegre, where we participated in the International Seminar "Disinformation, Social Media Platforms and Climate Crises." The event brought together an international network of researchers working across multiple continents to explore the intersection of climate change, politics, technology, and platform governance.
Professor Fabio Giglietto presented collaborative research with Dr. Giada Marino and Bruna Paroni that explores an innovative approach to analyzing political discourse: using fine-tuned large language models (LLMs) to examine political talk in hyperpartisan Facebook groups. This methodology offers a powerful way to process large datasets while maintaining nuance across different languages — a crucial advantage for research that extends beyond English-language platforms and communities.
AoIR2025 – Ruptures: Three Studies Presented in Niterói (October 16–18)
Building on this momentum, the team traveled to Niterói, Rio de Janeiro, to present at “AoIR2025 – Ruptures” (in full: the 2025 Association of Internet Researchers Conference), where we shared three distinct but complementary research studies.
Mapping Polarization and Political Narratives in Brazilian Facebook Groups
The first presentation was on Thursday 16 October, entitled "LLM-in-the-Loop for Identifying Narratives in Partisan Spaces on Facebook in Brazil". It showcased joint work by Fabio Giglietto, Bruna Paroni, and Giada Marino.
The research develops a methodological workflow to investigate the relationship between emotional polarization and political actors cited in Facebook posts from pro-Bolsonaro groups. A key innovation of this approach is its use of LLMs to recognize political actors in text while capturing the cultural and linguistic nuance of the discourse — including idiomatic expressions, cultural references, and sarcasm. This capability proves essential for accurately studying political communication in contexts beyond the English-speaking world.
Understanding How Deception Evolves Online
On the same day, Giada Marino presented "Throwing Spaghetti, Seeing What Sticks: Iterative Deception in Digital Strategic Information Operations," a collaborative study with Fabio Giglietto, Anwesha Chakraborty, Massimo Terenzi (Università degli Studi di Urbino Carlo Bo), and Samuel Olaniran (University of the Witwatersrand). The research explores a critical but often overlooked aspect of information warfare: how deceptive operations evolve through cycles of trial and adaptation. Rather than following a single, predetermined script, these operations continuously test which narratives and tactics resonate with audiences, refining their approach based on what "sticks." The work is based on broader research conducted through VERAAI_ALERTS, a tool developed within vera.ai to continuously monitor lists of problematic actors on Facebook and thus discover new deceptive information operations.
Navigating Data Access in the DSA Era
On Friday, 17 October, Fabio Giglietto moderated a roundtable discussion on "Ruptures in Data Access: The State of Social Media Research APIs & Tools in the DSA Era". It brought together leading researchers including Axel Bruns, Josephine Lukito, Jessica Yarin Robinson, and Richard Rogers (University of Amsterdam, also a vera.ai partner). The discussion focused on Articles 40.4 and 40.12 of the Digital Services Act, which establish two distinct pathways for researchers to access platform data — a critical conversation as the regulatory landscape for social media research continues to shift.
All these presentations marked an important stage in vera.ai, as - in the project - a lot of work focussed on integrating computational methods and social science approaches to study disinformation at scale. The research spans topics such as iterative deception in information operations and culturally informed narrative analysis, developing multilingual tools for examining digital manipulation across different contexts.
As the regulatory landscape evolves and information operations become more complex, vera.ai’s findings aim to support the research community and inform efforts to address disinformation. Our work, conducted at the University of Urbino, contributes significantly to all this.
Authors: Giada Marino, Fabio Giglietto (Uni Urbino)
Editor: Jochen Spangenberg (DW)
Images: Bruna Paroni (Uni Urbino)
