Conformidade ética da Inteligência artificial no Judiciário: diagnóstico empírico da ELIS do TJPE.

Abstract

Since 2020, the Brazilian Judiciary has regulated the use of artificial intelligence tools in the justice system, encouraging the implementation of such mechanisms in compliance with the ethical parameters established in Resolution No. 332/2020 of the National Council of Justice (CNJ). However, prior to the validity of said resolution, there were already artificial intelligence tools in use in the Brazilian judiciary, such as the ELIS tool used by the Court of Justice of Pernambuco (TJPE) to screen municipal tax executions. Thus, this Master's dissertation investigates the compliance of the ELIS artificial intelligence tool with the ethical principles established in multilateral, international, and national guidelines related to the tools used in the justice system and which were introduced in the national scenario through Resolution No. 332/2020 of the National Council of Justice (CNJ) and the recent Resolution No. 615/2025. The research combined theoretical and empirical methods. The bibliographic review, supported by national and international interdisciplinary literature (legal and computational), explored the concept of artificial intelligence (AI), its learning techniques, and the ethical guidelines applicable to the Brazilian judiciary, converging principle-based and abstract guidelines into objective practical commands and constructing a mapping of these essential ethical requirements for AI tools used in the Judiciary. The following ethical precepts were examined: autonomy and human dignity, equality (equity and nondiscrimination), transparency (publicity and explainability), security, and responsibility (accountability). Semi-structured interviews were conducted with developers, internal users of TJPE, and jurisdictions impacted by the ELIS tool, enabling a qualitative analysis of the research problem. The empirical analysis of ELIS, in addition to verifying compliance with ethical commands, also identified different practices regarding the use of the tool between the Districts of Recife and Olinda, revealing challenges and opportunities for ethical compliance. Finally, this dissertation presents empirical and theoretically based conclusions on the compliance of ELIS with ethical principles, including recommendations to TJPE and CNJ for the implementation of measures that guarantee ethics and security in the use of AI ELIS in the Judiciary.