Notions of Fairness in Automated Decision Making: An Interdisciplinary Approach to Open Issues

Artificial Intelligence (AI) systems share complex characteristics including opacity, that often do not allow for transparent reasoning behind a given decision. As the use of Machine Leaning (ML) systems is exponentially increasing in decision-making contexts, not being able to understand why and how decisions were made, raises concerns regarding possible discriminatory outcomes that are not in line with the shared fundamental values. However, mitigating (human) discrimination through the application of the concept of fairness in ML systems leaves room for further studies in the field. This work gives an overview of the problem of discrimination in Automated Decision-Making (ADM) and assesses the existing literature for possible legal and technical solutions to defining fairness in ML systems.