[{"jcr:title":"PIBIC research evaluates the autonomy of Meta’s Oversight Board","cq:tags_0":"tipos-de-conteudo:pesquisa-na-graduação","cq:tags_1":"area-de-conhecimento:direito","cq:tags_2":"tipos-de-conteudo:acontece-no-insper/programa-de-bolsas"},{"richText":"Law student Ruth Gaudêncio da Silva used Natural Language Processing techniques to classify public comments and analyze their impact on the board’s decisions","authorDate":"16/04/2026 13h47","author":"Tiago Cordeiro","madeBy":"Por","tag":"tipos-de-conteudo:pesquisa-na-graduação","title":"PIBIC research evaluates the autonomy of Meta’s Oversight Board","variant":"imagecolor"},{"jcr:title":"transparente - turquesa - vermelho"},{"themeName":"transparente - turquesa - vermelho"},{"containerType":"containerTwo"},{"jcr:title":"Grid Container Section","layout":"responsiveGrid"},{"text":"In 2018, Facebook CEO (now Meta) Mark Zuckerberg announced the creation of a new area within the company: the Oversight Board, which began operating at the end of 2020. According to the company’s official website, its role is to “help answer some of the most difficult questions around online freedom of expression: what to remove, what to leave up, and why.” The body has its own charter, internal regulations, and proposes to operate independently. Cases are selected by board members after users exhaust Meta’s appeals process, whether regarding their own content or content they have reported. The company itself may also refer cases for review. “People can appeal if they disagree with Meta’s content moderation decisions on Facebook, Instagram, or Threads. We commit to implementing the board’s content decisions, and its recommendations will help shape how we handle our policies,” the company states. But in practice, does public participation — expressed through comments — align with the board’s final decisions? The research project developed by Law student Ruth Gaudêncio da Silva sought to answer this question. The study, Judicial Behavior and the Facebook Oversight Board: The Role of Public Comments, was conducted as part of the Institutional Scientific Initiation Scholarship Program (PIBIC). As stated in the final report, it found that “public participation is highly concentrated in a small number of high-profile cases, while most cases receive limited engagement, which restricts the representativeness of civil society in the process.” On the other hand, the study also concluded that the correlation between the positions expressed in public comments and the board’s final decisions is limited, suggesting a certain degree of autonomy in its deliberations. A long-term project “If a post is interpreted as offensive, it is removed automatically or after being reported. But the author of the message can appeal internally. It is a way for Meta to reduce judicialization involving the use of its platforms,” says the student. “There are cases in which a post about breast cancer is removed for allegedly exposing the human body, or situations in which a certain word has a local meaning different from the one interpreted by those reporting it,” she explains. In such cases, the board acts as a court of last resort. According to Associate Professor Ivar Hartmann, the student’s work is part of a broader context. “Her project was inserted into a larger research agenda on the Oversight Board involving different people. This also allowed Ruth to reflect on issues of compartmentalization and on the relationship between stages and elements within the same research project.” The main objective — building a large database on all cases judged by the Oversight Board, containing dozens of variables for each case — was achieved and will support future research, he explains. “The project is important for understanding alternatives for online content moderation on social networks today. It is ongoing, with two other undergraduate students now working on a different stage of the research.” The methodology used by Ruth relied on a mixed approach, combining initial manual analysis to define classification parameters and, later, Natural Language Processing (NLP) techniques, using Large Language Models (LLMs) implemented in Python and the ChatGPT tool. “The automated analysis aimed to classify and analyze the content of comments, enabling the identification of patterns among them. The results showed a 90% agreement rate between automated and manual classification, demonstrating the potential of using LLMs to further understand the influence of public participation on the decisions of the Oversight Board,” the research report states. In this case, having studied programming in the Law program proved essential, says the student. “I already had the basic knowledge to get started. I also went through Insper Data, an experience that helped me a lot at this stage,” she explains, referring to the student organization focused on applying data science and data engineering to real-world problems.   Demystifying undergraduate research Professor Ivar highlights the benefits of developing undergraduate research projects. “First, direct exposure to the planning, execution, and reporting of studies conducted with scientific rigor. Second, training in skills related to research but also essential to any career, such as data analysis, critical thinking, and decision-making in unexpected situations. Third, the opportunity to build connections and strengthen relationships with faculty and students beyond one’s own undergraduate cohort. Fourth, understanding the process of scientific knowledge production, which is crucial for consuming news in today’s world.” Ruth, who is currently in the eighth semester of the Law program, is a full scholarship student and participated in the Digital Policy Leadership Program at the University of St. Gallen, in Switzerland, at the beginning of 2025. She says she experienced these benefits firsthand. “It’s a very valuable opportunity. But at the beginning of my undergraduate studies, I was unsure. I entered the program focused on the job market, concerned with activities that would help me secure internships more quickly. I had the impression that research was exclusively for those pursuing an academic career.” Once she started the project, however, she realized in practice the positive impact the experience could have on her trajectory. “The project gave me autonomy and time to work. It’s very different from classroom assignments, which are usually more guided by professors. In PIBIC, we explore as we develop the topic. In addition, working with research opens doors to master’s and doctoral programs, which are important for those pursuing professional — and not necessarily academic — careers in Law.”"}]