Análise dos enquadramentos de ação coletiva do movimento antidemocrático do 8 de janeiro de 2023 no Twitter/X através de métodos computacionais

Abstract

This dissertation investigates the collective action frames mobilized on Twitter/X by supporters of the coup-oriented movement during the antidemocratic events of January 8, 2023. Drawing on social movement theory and framing theory, alongside Computational Social Science methods, the research analyzes 1,337,606 tweets from the Interfaces Twitter Elections Dataset (ITED-Br), segmented into two temporal windows: the days preceding (January 5–7) and following (January 8–10) the events. Automatic labeling was performed using BERT models trained on 4,000 manually annotated tweets. The analysis combines stance classification, lexical frequency, co-occurrence networks, and bigrams. Results show that the pre-event period was dominated by prognostic frames, expressed through terms such as "war," "general strike," and "armed forces”, reflecting a discourse oriented toward direct action. In the aftermath, a shift toward diagnostic frames emerged, characterized by responsibility-deflection narratives ("infiltrators," "peaceful protesters") and antagonization of the establishment. Across both periods, collective action frames indicate a populist master frame grounded in the antagonism between "the people" and political institutions, with nationalism as a central vehicle for collective identity. A significant inversion in the proportion of pro and anti-coup tweets was also observed after the events, signaling public repudiation in the digital sphere. The work makes publicly available a replicable methodological pipeline and two original labeled datasets, advancing the dialogue between Political Science and Computational Social Science in the study of Brazil's democratic crisis.