ALGORITHMIC DISCRIMINATION IN ELECTRONIC HIRING PROCESSES AND A METHODOLOGY FOR THE ELIMINATION OF DISCRIMINATORY BIASES
Due to technological advancements, companies are increasingly adopting Artificial Intelligence (AI) systems in their departments, including human resources (HR) and recruitment and selection. However, it is observed that the risks of violating human rights and fundamental principles are growing, primarily through algorithmic discrimination, where biases and stereotypes are replicated. The methodology employed in this thesis involved reviewing technical articles in engineering and computer science, along with comparative legal analysis encompassing Brazilian legislation and the more advanced European Union regulations in this field. Additionally, case studies were conducted, examining the technical and legal aspects of discriminatory biases and stereotypes considering the Brazilian General Data Protection Law (LGPD) and the General Data Protection Regulation (GDPR) of the European Union, with interdisciplinary dialogue with other fields such as philosophy. A recruitment and selection algorithm was developed, and statistical metrics related to discrimination were evaluated. Statistical tests were also conducted to analyze statistical significance of differences between groups. Based on the results and analyses obtained, a four-phase methodology was proposed to eliminate discriminatory biases in electronic selection processes. This methodology includes implementing labor compliance tools to eliminate historical biases, utilizing algorithms that do not handle sensitive data, validating non-discriminatory algorithms through statistical metrics, and adopting affirmative action when necessary.