Inteligência artificial e direito: o impacto dos vieses algoritmicos nas decisões judiciais automatizadas pela tecnologia de aprendizado de máquina denominada machine learning

Abstract

This paper analyzes the impacts of algorithmic biases used in the Artificial Intelligence (AI) tools within the Judiciary, which generally seek to identify patterns from data analysis through mathematical logic (algorithm) and machine learning. - a machine learning. The present work defined algorithms, and their context in the Artificial Intelligence tools. The main object of the analysis was the vulnerability provided by the biases (discriminatory, prejudiced, biased ...) arising from the data used as a training parameter of AI systems, which will perform the analysis of judicial demands. To enrich the research, we report on US experience in the use of AI in Criminal Justice through the COMPAS system. The bibliographic research concluded, based on the experience of the American model, that the bias algorithms will provide great challenge in research, studies and regulation, in order to mitigate the biased results of injustices and prejudices within the scope of technological law.