Avaliação de impacto algorítmico (AIA): ferramenta para accountability, uso e desenvolvimento ético de IA pelo poder judiciário

Abstract

This paper investigates the development of governance and algorithmic accountability mechanisms for artificial intelligence (AI) systems in the Brazilian Judiciary, proposing an Algorithmic Impact Assessment (AIA) methodology for the use of AI development in the judicial context. The study is based on the observation of the growing use and development of AI systems in the Judiciary and the recently approved Resolution 615 of the National Council of Justice (CNJ), which represents a paradigmatic transformation in the administration of justice. The research addresses the risks of algorithmic discrimination, decision-making opacity, and violation of fundamental rights, especially regarding vulnerable groups. Using a qualitative and interdisciplinary methodology, it combines normative and documentary analysis with a comparative study of international regulatory frameworks. The theoretical framework articulates two perspectives: critical theory of technology (Crawford, Zuboff, Eubanks) and democratic accountability theory (Bovens), proposing a methodology for assessing algorithmic impact (Mantelero). The research is structured around four axes: mapping AI systems in the judiciary; analysis of ethical and legal risks; construction of accountability bases; and development of an AIA methodology. It is argued that the current mechanisms established by CNJ Resolution No. 615/2025, although representing progress, are insufficient to ensure effective accountability. As a main contribution, an AIA methodology is proposed that integrates technical, legal, and social safeguards, aiming to ensure transparency, prevent discrimination, and protect fundamental rights, balancing technological innovation with the constitutional values of the Brazilian justice system.