Artificial Intelligence in the Superior Court of Justice
Brazil's Superior Court of Justice, the STJ (Superior Tribunal de Justiça, the highest court on non-constitutional matters), has been actively engaged in developing and implementing AI solutions since 2018, when it adopted its "pilot project for the application of artificial intelligence solutions" to speed up the processing of cases in the court by "increasing work productivity and efficacy" and "automating and streamlining work routines".
Under the project, appeals are automatically classified through a screening process that involves textual analysis of appellants' briefs to extract the provisions of law on which the appeal is based (Instruction IN 6/2018).
The year 2018 also saw the creation of a new tool for case law research, the "Corpus927 System" (Sistema Corpus927), offering various functionalities, including (i) grouping binding precedents, (ii) automatic extraction of references to legislation in each decision, (iii) analysis of similarities in the decisions, (iv) grouping similar decisions, with an indication of the most common positions, based on the percentage of similarities, and (v) the creation of a real-time decision database sourced from each court's website (Joint Directive PC 1/2018).
In 2021, the STJ established its "Managing Committee for Artificial Intelligence Ethics", with the mission of assessing and approving algorithms and AI models used by the court, identifying risks associated with use of AI models, and recommending security and control measures (Directive P 365/2021). In 2022, the STJ introduced a "Laboratory for Innovation and Sustainable Development Objectives" to facilitate solutions and innovations for the court's judicial and administrative routines (Directive P 12/2022).
The STJ currently has an Artificial Intelligence Advisory (AIA – Assessoria de Inteligência Artificial) linked to the office of the STJ's President, focused on enhancing "productivity in judicial work", along with an Information Technology and Communication Governance Committee, an Information Technology Managing Committee, and Managing Committees for Administrative and Judicial Systems (Instruction IN 15/2023).
Artificial intelligence is part of the STJ's 2023/2024 Strategic Plan for Information Technology and Communications. The plan includes the creation of a platform to extract and send procedural data to the National Council of Justice (CNJ – Conselho Nacional de Justiça), and features two AI-based initiatives among its strategic projects: "Advanced Procedural Screening", aimed at "streamlining the processing of cases and reducing the volume of work in justices' chambers", and "AI-supported Admissibility Analysis", which is designed to identify the grounds relied on by the lower court to deny appeals to the STJ and correlate them with the arguments contesting those grounds in the appellant's interlocutory appeal, "in order to reduce the time spent by the Office of the President of the STJ on documenting and analyzing admissibility of appeals".
Currently, the STJ employs two AI solutions, the Athos System and the Socrates Project. The Athos System, which was developed by the STJ and has been in use since 2019, monitors legal issues appearing frequently in litigation, with a view to "intensifying the formation of binding precedents" by identifying "possible new paradigms for selecting issues" to be decided under the repetitive appeal system. In addition, Athos identifies, among the cases before the STJ, those that raise issues identical to matters on which the STJ has issued (or will issue) a binding ruling in a repetitive appeal proceeding1. Athos can also identify (i) cases in which decisions by the STJ's panels and sections have made converging or diverging decisions, (ii) cases involving highly significant issues and (iii) cases in which binding precedents were distinguished or not followed.
The Athos System is used by the Admissibility, Repetitive Appeals and Important Issues Advisory (ARP – Assessoria de Admissibilidade, Recursos Repetitivos e Relevância) attached to the Office of the STJ's President. The ARP is responsible for overseeing the admissibility of appeals and proceedings in which the STJ has original jurisdiction, compliance with the repetitive appeals system, and the importance of the issue of federal law raised on appeal. Most cases in the STJ are first screened by the APR, where one of the steps is to determine whether the case raises issues decided (or to be decided) under the STJ's repetitive appeal system.
Associated with the ARP, the Precedents Management Center (NUGEPNAC – Núcleo de Gerenciamento de Precedentes) is responsible for managing the binding precedents systems provided for under the Code of Civil Procedure of 2015. Among other tasks, the Center identifies potentially repetitive issues in the cases that come before the court to facilitate the assignment of cases to the repetitive appeals system. The Athos System plays a fundamental role in this work as well.
According to Justice Sanseverino, the system "seeks to give priority to deciding legal issues, stating the STJ's interpretation of the law in summary form with respect to the question at issue in the appeal, not resolving individual cases", thus supporting the STJ's institutional mission of ensuring uniformity in the interpretation and application of federal legislation, ensuring that the primary focus is on the legal issues, with the "satisfaction of the parties' interests" coming second2.
The plan is to expand the system to the 32 courts under the STJ's jurisdiction, an initiative known as "Athos Courts" (Athos Tribunais), whose goal is to enhance the formation of binding precedents.
The STJ also employs the Socrates Project, in use since 2019. The rationale behind the project is the same as the Athos System's, but Socrates is used directly by the Justices' chambers, to "reduce time and effort spent in screening cases, support review of proceedings, and assist in selecting representative cases on disputed issues." The tool also has the capacity to review the proceedings in the chamber to identify those that have common issues, compare pleadings, and review the court's decisions to identify relevant precedents3.
Socrates has been improved and updated and the current version, Socrates 2.0, can analyze pleadings in appeals to the STJ, indicate the constitutional provision on which the appeal is based, the legislative provisions at issue in the appeal, and the cases cited to demonstrate a divergence in the precedents. It can also generate "word clouds", highlighting the most important words in appeal briefs. After validation by users, Socrates will indicate the items of the appeal that are potentially inadmissible, thus facilitating decision-making.
In this rapidly evolving landscape, the STJ is at the forefront of AI use by Brazil's courts. Through technical cooperation agreements with state appeal courts, federal appeal courts and the federal Public Prosecution Service, the STJ is working to improve data flow, streamline court procedure, and modernize the judicial system.
NOTES
1 FIGUEIREDO, Guilherme Silva. PROJETO ATHOS: Um Estudo de Caso sobre a inserção do Superior Tribunal de Justiça na Era da Inteligência Artificial. Dissertação de Mestrado; Universidade de Brasília, 2022.
2 SANSEVERINO, Paulo de Tarso; MARCHIORI, Marcelo Ornelas. O projeto Athos de inteligência artificial e o impacto na formação de precedentes qualificados no Superior Tribunal de Justiça. In: O Direito Civil na era da Inteligência Artificial / Rodrigo da Guia Silva e Gustavo Tepedino (Coordenadores). 1. ed. São Paulo: Thomson Reuters Brasil, 2020.
3 NORONHA, João Otávio de. Os avanços da inteligência artificial no STJ: caminho para uma Justiça mais rápida e efetiva. In: O Direito Civil na era da Inteligência Artificial / Rodrigo da Guia Silva e Gustavo Tepedino (Coordenadores). 1. ed. São Paulo: Thomson Reuters Brasil, 2020.
