Generative Artificial Intelligence in the Judiciary:

training and impact of legal writing whit ChatGPT in the 5th Region

Authors

DOI:

https://doi.org/10.56256/themis.v23i1.1108

Keywords:

Treinamento e Aperfeiçoamento de Juízes, Pedagogia, Atividade Jurisdicional, Inteligência Artificial Generativa, ChatGPT

Abstract

The course "Legal Writing with ChatGPT: Theory and Practice" was designed to train judges to use artificial intelligence (AI) in their legal writing practice through ChatGPT, aiming to enhance the effectiveness and persuasiveness of their texts. One year after the course, an evaluation was conducted to analyze the impact of Generative AI on the judicial activities of judges in the 5th Region. The objective was to assess how the course influenced the efficiency and quality of legal document production. The study uses a mixed-methods approach, both qualitative and quantitative, including data collection via the Teams Platform, with questions about institutional affiliation, the use of ChatGPT, and its impact on judicial activities. The research is exploratory and descriptive in its objectives and strictly analytical in nature. The data showed that 71.42% of participants noticed a reduction in writing time, and 81.48% observed improvements in document quality. Generative AI was predominantly used for text revision and decision structuring. Despite significant use of the tool, only 30% of judges' work teams utilized it. Suggestions for improvement included practical integration of AI into the work environment, enhancements in educational materials, and new courses on AI innovations. In conclusion, Generative AI proved effective in optimizing time and improving the quality of legal texts, with potential for positively impacting judicial productivity.

Author Biographies

José Eduardo de Melo Vilar Filho, Universidade Federal do Ceará - UFC

Doutorando em Direito (UFC), Mestre em Direito (UFC/2006). Pesquisador do PPGD/ENFAM (Escola Nacional de Formação de Magistrados). Diretor da Escola da Magistratura Federal no Ceará (ESMAFE/CE). Juiz Federal. Fortaleza, CE, BR. E-mail: <eduardo.vilar.filho@gmail.com>.

George Marmelstein, Centro Universitário 7 de Setembro (UNI7)

Professor do Centro Universitário 7 de Setembro (UNI7), lecionando na graduação e no mestrado. Doutor em Direito, Justiça e Cidadania no Século XXI pela Universidade de Coimbra. Mestre e Graduado em Direito pela UFC. Professor orientador doNúclero de Pesquisa em Interpretação e Decisão Judicial (NUPID). Juiz Federal. Fortaleza, CE, BR. E-mail: <georgemlima@yahoo.com.br>.

Carlos Eduardo Ferreira Aguiar, Universidade Federal do Ceará - UFC

Doutorando e Mestre em Direito pela Universidade Federal do Ceará (UFC). Graduado em Direito pela Universidade Estadual Vale do Acaraú (UVA). Assistente Técnico da Escola de Magistratura Federal Núcleo Seccional do Ceará (ESMAFE-CE). Fortaleza, CE, BR. E-mail: <car.guiar.18@gmail.com>.

Published

2025-03-26

How to Cite

Vilar Filho, J. E. de M., Marmelstein, G., & Aguiar, C. E. F. (2025). Generative Artificial Intelligence in the Judiciary:: training and impact of legal writing whit ChatGPT in the 5th Region. THEMIS: Revista Da Esmec, 23(1), 145–170. https://doi.org/10.56256/themis.v23i1.1108

Issue

Section

ARTIGOS CIENTÍFICOS