DECISIONAL TRANSPARENCY IN THE ALGORITHMIC ERA

EXPLAINABLE ARTIFICIAL INTELLIGENCE AS A REQUIREMENT FOR JUDICIAL REASONING

Abstract

This article investigates the tension between the increasing adoption of artificial intelligence systems by the Brazilian Judiciary and the constitutional principle of reasoned judicial decisions. The research is justified by the rapid incorporation of algorithmic technologies into judicial services and the risks they pose to decision-making rationale. The main objective is to assess whether the legitimacy of algorithm-assisted decisions depends on the explainability of their outputs and to propose criteria for reconciling technological innovation with fundamental procedural guarantees. Methodologically, qualitative theoretical-documentary research is conducted through normative, jurisprudential, and bibliographic analysis. The study examines CNJ Resolutions No. 332/2020 and 615/2025, the General Data Protection Law, Bill No. 2,338/2023, the Justice in Numbers 2024 Report, UNESCO Recommendation on AI Ethics, precedents from Brazilian Supreme Courts, comparative analysis of State v. Loomis, and specialized literature on judicial reasoning, algorithmic bias, and explainable AI. Results demonstrate that algorithmic opacity and discrimination undermine the integrity, dialogical nature, correctness, and rationality of judicial reasoning. Risks and failure modes pertinent to AI use in judicial decisions are systematized. Minimum guidelines are proposed: local and global system explanations, audit trails, judicious use of explainability techniques, meaningful human oversight, and contestation procedures. The conclusion establishes that algorithmic explainability constitutes a constitutional imperative to ensure effective adversarial process and preserve the democratic legitimacy of the judicial function.