Entre a opacidade e a explicabilidade

• Autor(a) principal: Negruni, Lisiane Penz
• Data de Publicação: 2023
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10362/153333
• topic: Administração pública; Opacidade; Explicabilidade; Decisões algorítmicas; Motivação; Direito à explicação; Public administration; Opacity; Explainability; Algorithmic decision-making; Right to explanation; Direito

The article aims to establish the new contours for the duty of motivation of individual administrative acts, considering the scenario of digitalization of the Public Administration and the use of algorithms to produce administrative decisions. From the legal content of the duty of motivation, as a corollary of the principle of legality, it is argued that its observance, in a context of algorithmic administrative decisions, consists in ensuring that the citizen is able to understand and challenge the decision-making process, or, in other words, counterargument the algorithm. It is shown that, technically, there are means to ensure a higher level of explicability to the results provided by algorithmic systems and that, legally, a right to explanation stems from a global interpretation of the LGPD. It is concluded that, in order to make use of algorithmic administrative decisions, the Public Administration should safeguard explainability as a requirement of the motivation, in order to determine that the choice of algorithm observes demanding standards of accessibility and interpretability, which imposes a careful look at the human choices made at all stages of design and implementation of the model (explainability by design). It is recommended the regulation of the topic in the Brazilian legal system, especially in face of the normative uncertainty regarding the types of administrative decisions that can be automated, the degree of automation appropriate in each case and the lack of clarity on the need for human intervention in the review process of these decisions.