Supervised learning, explanation and interpretation from pretrial detention decisions by Italian and Brazilian Supreme Courts
Loading...
License
Cadmus Permanent Link
Full-text via DOI
ISSN
0302-9743; 1611-3349
Issue Date
Type of Publication
Keyword(s)
LC Subject Heading
Other Topic(s)
EUI Research Cluster(s)
Initial version
Published version
Succeeding version
Preceding version
Published version part
Earlier different version
Initial format
Citation
Tiago PRINCE SALES, João ARAÚJO, José BORBINHA and Giancarlo GUIZZARDI (eds), Advances in conceptual modeling, Cham : Springer, 2023, Lecture notes in computer science ; 14319, pp. 131-140
Cite
BILLI, Marco, RAULINO DAL PONT, Thiago, SABO, Isabela Cristina, LAGIOIA, Francesca, SARTOR, Giovanni, ROVER, Aires José, Supervised learning, explanation and interpretation from pretrial detention decisions by Italian and Brazilian Supreme Courts, in Tiago PRINCE SALES, João ARAÚJO, José BORBINHA and Giancarlo GUIZZARDI (eds), Advances in conceptual modeling, Cham : Springer, 2023, Lecture notes in computer science ; 14319, pp. 131-140 - https://hdl.handle.net/1814/78104
Abstract
Pre-trial detention is a debated measure in different legal systems since it deprives defendants of their liberty prior at the initial stage of proceedings. To order this measure, the judge must justify it by highlighting the risks that the arrested person presents to society and to the criminal procedure itself. An example of a factor related to preventive custody, in countries such as Italy and Brazil, is involvement in criminal organizations. The paper presents the results of experimental research with supervised learning, in particular using XAI techniques, such as decision trees and Shapley Additive Explanations. Our corpora are composed of unstructured data (texts of judicial decisions) and structured data (factors extracted from such judicial decisions), from the case law of Italian and Brazilian Supreme Courts. As a result, we have identified a collection of factors that play an important role in the reasoning of the judge and in predicting outcomes, including common factors between the two countries. In particular, we have verified that involvement in criminal organizations consistently leads to the decision to maintain imprisonment in the Brazilian scenario, while in the Italian context, this is unclear. Finally, we conclude that data structuring based on the extraction of factors from the decision texts not only increases the prediction’s quality but also allows for their interpretation and explanation.
Table of Contents
Additional Information
Published online: 16 October 2023
External Links
Publisher
Version
Research Projects
European Commission, 833647
Sponsorship and Funder Information
This research has been supported by Brazilian National Council for Scientific and Technological Development (CNPq) and Coordination for the Improvement of Higher Education Personnel - Institutional Program for Internationalisation (CAPES/PrInt); ADELE (Analytics for Decision of Legal Cases, EU Justice program Grant (2014–2020); COMPULAW (Computable law G.A. 833647), ERC Advanced Grant (2019-2024); LAILA (Legal Analytics for Italian Law), MIUR PRIN Programme (2017), the European Commission under the NextGeneration EU programme, PNRR - M4C2 - Investimento 1.3, Partenariato Esteso PE00000013 - “FAIR - Future Artificial Intelligence Research” - Spoke 8 “Pervasive AI”.