Algorithmic bias: a provisional review

• Autor(a) principal: Simões-Gomes, Letícia
• Data de Publicação: 2020
• Outros Autores: Roberto, Enrico, Mendonça, Jônatas
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Estudos de Sociologia
• Texto Completo: https://periodicos.fclar.unesp.br/estudos/article/view/13402
• topic: Algorithmic bias; Algorithm; Discrimination; Digital Sociology; Sesgo algorítmico; Algoritmo; Discriminación; Sociología digital; Viés algorítmico; Algoritmo; Discriminação; Sociologia Digital

This paper stands in the field of Digital Sociology and proposes to carry out a provisional bibliographic review on algorithmic bias, based on the main journals databases in English, Spanish and Portuguese. Our aim is to assess how the debate on algorithmic bias in the Humanities has developed, observing which definitions, origins, diagnostics and perspectives are presented for the phenomenon. It is observed that the majority of articles are of essayistic nature and produced in the Global North, with a low penetration of the subject in Spanish and Portuguese language literature. The expression remains broadly undefined, and is sometimes treated as synonymous with discrimination, sometimes as its cause. The main sources of bias are, according to the literature, the model development and its training data, which largely lead to recommendations for increasing transparency about the process and suggest an analytical tendency to emphasize the subjective character of algorithmic bias. There seems to be an analytical tendency in the subjective character of bias. As a perspective, we point out the importance of integrating the elements transcending the subjectivity of these actors to this analysis.