Comportamento Judicial e o Facebook Oversight Board: O papel dos comentários públicos
Autores
Silva, Ruth Gaudêncio da
Co-orientadores
Citações na Scopus
Tipo de documento
Relatório de Iniciação Científica
Data
2025
Resumo
Esta pesquisa investiga a influência dos comentários públicos de todas as decisões
do Oversight Board do Facebook até o primeiro quadrimestre de 2025, buscando
entender como a quantidade, a origem e o posicionamento desses comentários se
relacionam com o resultado dos casos analisados. O objetivo é determinar se a
participação do público, expressa por meio de comentários, converge ou diverge das
decisões finais do conselho. A metodologia empregada foi uma abordagem mista,
combinando análise manual inicial para definir os parâmetros de classificação, e
posteriormente, com o uso de técnicas de Processamento de Linguagem Natural
(PLN) e aprendizado de máquina, com o uso de Large Language Models (LLMs)
implementados em Python e da ferramenta ChatGPT. A análise automatizada visa
classificar e analisar o conteúdo dos comentários, permitindo a identificação de
padrões nos comentários. Os resultados indicaram taxa de concordância entre a
classificação automatizada e a manual, de 90%, demonstrando o potencial da
utilização de LLMs para compreender, posteriormente, a influência da participação
pública nas decisões do Oversight Board. A análise revelou que a participação pública
é altamente concentrada em poucos casos de grande repercussão, enquanto a
maioria recebe baixo engajamento, o que limita a representatividade da sociedade
civil no processo. Além disso, observou-se a sub-representação do Sul Global,
evidenciando desigualdades regionais na participação. Por fim, verificou-se que a
correlação entre o posicionamento dos comentários públicos e as decisões finais do
conselho é limitada, o que sugere relativa autonomia do Oversight Board em suas
deliberações. Desse modo, estudo contribui para a compreensão do funcionamento e
da eficácia do conselho como um mecanismo de governança, com autonomia e
legitimidade das suas decisões.
This paper investigates the influence of public comments on all decisions of Facebook’s Oversight Board up to the first quarter of 2025, seeking to understand how the quantity, origin, and positioning of these comments relate to the outcomes of the analyzed cases. The objective is to determine whether public participation, expressed through comments, converges with or diverges from the Board’s final decisions. The methodology employed was a mixed approach, combining an initial manual analysis to define classification parameters and, subsequently, the use of Natural Language Processing (NLP) and machine learning techniques, through the application of Large Language Models (LLMs) implemented in Python and the ChatGPT tool. The automated analysis aimed to classify and examine the content of the comments, enabling the identification of patterns. The results indicated a 90% agreement rate between the automated and manual classifications, demonstrating the potential of LLMs to further explore the influence of public participation on Oversight Board decisions. The analysis revealed that public engagement is highly concentrated in a few high-profile cases, while the majority receive low levels of participation, limiting the representativeness of civil society in the process. Moreover, the underrepresentation of the Global South was observed, highlighting regional inequalities in participation. Finally, the correlation between the stance of public comments and the Board’s final decisions was found to be limited, suggesting a relative autonomy of the Oversight Board in its deliberations. Thus, this study contributes to the understanding of the functioning and effectiveness of the Board as a governance mechanism, with autonomy and legitimacy in its decisions.
This paper investigates the influence of public comments on all decisions of Facebook’s Oversight Board up to the first quarter of 2025, seeking to understand how the quantity, origin, and positioning of these comments relate to the outcomes of the analyzed cases. The objective is to determine whether public participation, expressed through comments, converges with or diverges from the Board’s final decisions. The methodology employed was a mixed approach, combining an initial manual analysis to define classification parameters and, subsequently, the use of Natural Language Processing (NLP) and machine learning techniques, through the application of Large Language Models (LLMs) implemented in Python and the ChatGPT tool. The automated analysis aimed to classify and examine the content of the comments, enabling the identification of patterns. The results indicated a 90% agreement rate between the automated and manual classifications, demonstrating the potential of LLMs to further explore the influence of public participation on Oversight Board decisions. The analysis revealed that public engagement is highly concentrated in a few high-profile cases, while the majority receive low levels of participation, limiting the representativeness of civil society in the process. Moreover, the underrepresentation of the Global South was observed, highlighting regional inequalities in participation. Finally, the correlation between the stance of public comments and the Board’s final decisions was found to be limited, suggesting a relative autonomy of the Oversight Board in its deliberations. Thus, this study contributes to the understanding of the functioning and effectiveness of the Board as a governance mechanism, with autonomy and legitimacy in its decisions.
Palavras-chave
Plataformas digitais; Facebook Oversight Board; Decisão Judicial; Digital platforms; Facebook Oversight Board; Court Decision
Titulo de periódico
URL da fonte
Título de Livro
URL na Scopus
Sinopse
Objetivos de aprendizagem
Idioma
Português
Notas
Membros da banca
Área do Conhecimento CNPQ
CIENCIAS SOCIAIS APLICADAS
CIENCIAS SOCIAIS APLICADAS::DIREITO
CIENCIAS SOCIAIS APLICADAS::DIREITO
