Avanços iniciais do projeto concil-ia

desafios e potenciais impactos

Abstract

The article explores the initial progress of the Concil-IA project, developed as a response to the growing volume of cases in the Brazilian Judiciary. This scenario, which creates a significant workload, has prompted the National Council of Justice (CNJ) to implement policies for self-compositional solutions and technologies based on Artificial Intelligence (AI). The objective of the Concil-IA project is to create an AI-supported online dispute resolution (ODR) system, especially focused on consumer law cases, to facilitate conflict resolution and reduce litigation. The project’s methodology includes collecting judicial rulings, identifying key variables, and using the GPT model to automatically extract these variables, as well as developing a classification model to predict ranges of compensation for moral damages, with explainable predictions. The initial results show that the GPT-4.5 model achieved an accuracy rate of over 90% in the automated extraction of variables from judicial rulings, while the compensation classification model obtained an average accuracy of approximately 50%, indicating a need for further adjustments and improvements. Additionally, a web interface was developed to make the model accessible to mediators and parties involved in the conciliation process, offering a user-friendly platform that facilitates interaction with the predictive model and supports real-time decision-making. In conclusion, the progress made suggests that AI can innovate and make the judicial system more efficient, promoting more accessible and conciliation-oriented justice.