Discrimination in artificial intelligence

Abstract

Artificial Intelligence and algorithms have unlimited prospects to improve human lives but, at the same time, they pose unfathomed risks for human rights protection. Some of these risks have already manifested, as several discrimination claims, especially based on gender and race, have been brought to courts and media, questioning the platform economies using these technologies, and several experimental studies have shown their discriminatory patterns. It is relevant to examine the application of AI to this business model as the network effects it generates can exacerbate the discriminatory potential. This study analyzes how structural discriminations have perpetuated in Uber, one of the companies with more complaints, and explores the reasons behind this situation in order to determine whether the origin is vertical discrimination, due to the configuration of AI exclusively; the aggregation of horizontal discrimination by its users the algorithm enables, or a combination of both. It has been essential to understand the roots of discrimination in AI in order to develop an accurate account of relevant characteristics of the regulatory responses, self-regulation and legislation in the US and the EU, and determine whether they provide good prevention or an adequate response to the issue.