Inteligência artificial no poder judiciário: análise do risco de vieses algorítmicos em decisões judiciais

• Ano de defesa: 2025
• Autor(a) principal: Sampaio, Aline Bessa
• Orientador(a): Albuquerque, Felipe Braga
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Área do conhecimento CNPq: CNPQ::CIENCIAS SOCIAIS APLICADAS
• Link de acesso: http://repositorio.ufc.br/handle/riufc/80882

This dissertation aims to analyze the application of Artificial Intelligence (AI) systems in the Brazilian Judiciary, with an emphasis on identifying and mitigating algorithmic biases in judicial decisions. Considering the potential of these technologies to promote greater efficiency, procedural speed, and standardization of understandings, the research is based on the premise that their adoption, when carried out without rigorous technical and legal criteria, can compromise constitutional principles such as equity, impartiality, adversarial proceedings, and legal certainty. The methodology adopted is based on a bibliographic review and documentary analysis of standards, scientific articles, and practical cases. The work is structured in four chapters: the first presents the concepts and evolution of AI; the second analyzes its application in the Brazilian Judiciary; the third delves deeper into the study of algorithmic bias, discussing its sources, impacts, and risks; and the fourth proposes mitigation strategies, such as data governance, algorithmic auditing, and regulatory regulation. Among the main results, it is worth highlighting that biases often derive from biased historical data, which, when used to train algorithms, can perpetuate inequalities and compromise the legitimacy of automated decisions. In view of this, the importance of data curation, system explainability and human supervision throughout the entire AI life cycle is highlighted. It is also worth highlighting the need for solid and adaptive regulatory frameworks that establish minimum parameters of transparency, accountability and control over the systems used. The research concludes that the articulation between technological innovation and constitutional values must be guided by a clear public policy, transparent governance practices and ongoing training of the professionals involved. Thus, AI must be understood as an instrument to support jurisdictional activity, and not as a substitute for human action, ensuring that technology is used in an ethical, democratic manner and in line with fundamental rights.