Ranking decisions in Brazilian Supreme Court: a study of network of decisions robustness and leading cases identification

• Ano de defesa: 2023
• Autor(a) principal: Souza, Jackson José de
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Authority scores; Brazilian Supreme Court; Citation network; Decision network; Importance scores; Kleinberg; Natural Language Processing; PageRank; Pontuação de autoridade; Pontuação de importância; Processamento de Linguagem Natural; Rede de citações; Rede de decisões; STF; Supremo Tribunal Federal
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/45/45134/tde-12032025-105740/

This work studies quantitative measures for ranking judicial decisions by the Brazilian Supreme Court (STF) and selecting leading cases, understood as those with broadness of influence on different legal fields. The measures are based on a network built over decisions whose cases were finalized in the Brazilian Supreme Court between 01/2001 and 12/2019, obtained by crawling publicly available STF records. Three ranking measures are proposed; two are adaptations of the PageRank algorithm, and one adapts Kleinberg\'s Algorithm. All are compared with respect to the agreement on the top 100 rankings; we also analyze each robustness measure based on self-agreement under perturbation. We perform this analysis for two decision networks: the first one containing all collegiate decisions and the second one with some decisions filtered out guided by legal expertise on the works of the Supreme Court. In addition to comparisons made between algorithms on the same network, we also compared the results between both networks with respect to their ability to identify leading cases. We examine for each ranking whether the resulting quantitative ranking is congenial to a qualitative intuition of what the legal community usually considers as relevant precedents and discuss some possible criteria of relevance, trying to find out if there is any sense where the quantitative and the qualitative measures would better align. We conclude that after filtering low-relevance decision types, the STF decision network is still robust under 10%-perturbation, but presents higher degradation by increasing perturbation levels. Both versions of PageRank, even if producing different rankings, achieved robustness results that are indistinguishable by statistical tests. Kleinberg\'s algorithm provides a different ranking, with many relevant criminal cases in the top 100, according to legal experts consulted during this work. Relevant cases were ranked at the top due to the effectiveness of decision type filtering, which obtained a more meaningful and less noisy decision network.