Discriminação em algoritmos de inteligência artificial: uma análise acerca da LGPD como instrumento normativo mitigador de vieses discriminatórios

• Autor(a) principal: Ribeiro, Ana Lídia Lira
• Data de Publicação: 2021
• Tipo de documento: Trabalho de conclusão de curso
• Idioma: por
• Título da fonte: Repositório Institucional da Universidade Federal do Ceará (UFC)
• Texto Completo: http://www.repositorio.ufc.br/handle/riufc/57947
• topic: Discriminação; Algoritmos; Inteligência Artificial; LGPD; Decisões automatizadas

Investigate the existence of bias in artificial intelligence algorithms and the possibility of the LGPD being a normative instrument in order to mitigate the discriminatory bias in the decision-making of these systems. To do so, it was sought to, first of all, explain what is artificial intelligence, elucidating how its most important technologies work and the relevance of big data in this area. Afterwards, due to the polysemy of the word, it became necessary to delimit the approached concept of discriminination, along with an analysis of the existing normatives on the brazilian legal order that conceptualize it. After the initial concepts, it was explained what is an algorithm and, then, it was presented different examples of the occurrence of bias in the automated decision-making systems. Finally, it was conceptualized what is meant by algorithmic discrimination and in what ways an algorithm can become biased. Subsequently, the LGPD was analyzed about the data subject safeguards laid out in the legislation that help mitigate discriminatory biases in algorithms, and then the limits of these safeguards were looked at. The methodology used is exploratory and qualitative, based on bibliographic research, using articles, monographs, international documents, and master's dissertations, as well as documentary research, with articles of law and other normative acts. It was concluded that artificial intelligence algorithms are not neutral, since they can make discriminatory decisions through biases in their systems. Furthermore, the LGPD, besides instigating deeper discussions about the problematic approached so that, in the future, it will be possible to create a regulamentation focused on artificial intelligence, shows itself to be a good instrument to mitigate these biases to the extent that it brings indications legally imposed on developers and companies the obligation to produce deeper investigations and audits in order to analyze the true discriminatory potential of their systems and ensure results that are in compliance with the legislation.