Entre humanos e algoritmos

fundamentos, proposta e limites da petição machine friendly no processo judicial

Abstract

The incorporation of artificial intelligence systems into the Brazilian Judiciary has progressively reshaped the informational framework that precedes judicial adjudication. Although the decisional act remains vested in the human judge, the organization, triage, and, in certain procedural workflows, the synthesis of procedural records are increasingly mediated by computational systems employed as institutional support tools. In this context, legal pleadings no longer circulate exclusively within human cognitive processes, but rather traverse technical stages capable of influencing how a case is initially framed and presented to the adjudicator. This article examines the risks inherent to such algorithmic mediation, with particular attention to fidelity failures, structural omissions, and positional biases identified in the literature on natural language processing applied to long and complex legal documents. Based on this diagnosis, the article advances the concept of machine friendly pleadings as a technique for structuring legal discourse aimed at ensuring dual legibility—human and algorithmic—while preserving argumentative integrity. The proposal is grounded in a risk reduction approach and expressly rejects any notion of automating or manipulating judicial decision making.