Project description

In Argentina, all complaints that generate a case in the federal justice system are drawn through a computer program to determine the court that will investigate the case. This draw, according to the Judicial Power, should be “automatic and equal”, in order to guarantee an independent investigation of the interests of the parties involved and to minimize possible influences in the course of the investigation.

However, the randomness and independence of this lottery system has been frequently questioned. Cases such as the draw of the Nisman case and the raffles of judicial auctions generated suspicions about possible manipulations and irregularities in the computer system that assigns the cases to the courts. Due to these suspicions, Judge Servini de Cubría is currently investigating the computer system and its possible flaws.

This project seeks to analyze, through the official data of the judicial power published from July 2013 until December 2017, if the federal causes are truly distributed in an equitable and random manner among all the courts. Likewise, the project focuses on the distribution of causes of corruption, which are suspected to be more vulnerable to potential manipulations.

What makes this project innovative?

Argentina's judicial and federal adjudication software works since 2003, was always criticized by NGOs and journalists due to the hypothesis of manipulation. Furthermore, it was always difficult to prove the hypothesis due to lack of data. Judicial draw is a project to validate these hypotheses through data and statistical analysis. The project is innovative from the development and use of data that has 3 parts: 1 data acquisition through a scraper in the judicial site that reveals the causes from the Judicial Information Center  2- Statistical analysis and elaboration of a statistical model called "Monte Carlos". This model, through computational simulation, reveals the distribution of cases by court. As the distribution obtained through this method and the information provided by the Judicial Information Center, the manipulation of the system is demonstrated We also develop visualizations using Tableau and D3 to demonstrate that the system of judicial draws is manipulated and is not accurate as it should be this judicial system.

What was the impact of your project? How did you measure it?

The investigation was accompanied by journalists from La Nación in Argentina publishing an article with the investigative method and the Monte Carlo model to reveal the manipulation of the federal lottery system. This revelation is unique and demonstrates through the use of data science and data journalism the iregularities of the judicial system in Argentina. You can see more in this visual explanation about the method, in an interview with La Nación Data: http://blogs.lanacion.com.ar/data/entrevistas/data-science-franco-bellomo-pone-la-lupa- in-the-justice /

Source and methodology

We elaborated a statistical method to determine the most probable distribution, a mechanism called "Monte Carlo model". That showed that the system tends to a detour. "The system has a distribution method that does not give a uniform result, each year there are three or four courts that come out more favored, with a variation of 20% with respect to the one that has least." The favored court curve is moved every year. In the last four years it was concentrated from the middle to the end, "explained Franco Bellomo, data analyst and project member. We work with the data published on the website of the Judicial Information Center (http://www.cij.gov.ar/) about the draws of rooms of criminal cases in Federal Criminal and Correctional from July 2013 to December 2017 Through Machine Learning and data cleaning we were able to visualize the data in question. We have more than 32,900 drawings of judicial cases corresponding to the National Chamber of Criminal and Correctional Matters and almost 300 raffles for corruption cases. For this analysis, we only used the causes where the reason for allocation was by lot, discarding the causes where the assignment was due to a change of assignment or incompatibility.

Technologies Used

We develop scrapers: 1 - To normalizar la encuesta 3 - Scraper the Center of Juditial Information 4 - Analyze, order, structure, unify the data with Python 5 - We use Neo4j as a database 6 - Python-flask as backend 7 - we develop in JavaScript with vue.js to develop the front 8 - Tableau and d3.JS for the multiple dynamic visualizations We take care of releasing the datasets for use to contribute to the open data culture. Also the code is freely available to replicate.

Project members

Franco Bellomo Maia Jastreblansky Yas Garcia Lucia He Renzo Lavin

Link

Additional links

Followers

Click Follow to keep up with the evolution of this project:
you will receive a notification anytime the project leader updates the project page.