Читать книгу Data protection for the prevention of algorithmic discrimination - Alba Soriano Arnanz - Страница 30

Another classification is the one that differentiates between fully automated and semi-automated decision-making systems. While in the former there is no significant human intervention, the latter does involve some degree of human involvement and oversight, meaning the individual or group in charge of controlling the algorithm can overturn the decisions it makes. Semi-automated decision-making systems can also be used as tools for support in human decision-making processes. This classification is particularly relevant given the fact that more legal constraints are generally placed on fully automated decision-making systems.174 It is therefore important to analyse the differences in the risks that automated and semi-automated systems may pose on individuals for, as will be discussed later on, the assumption made by much of the scholarship and regulators that many of the risks generated by fully automated systems can be solved by introducing a “human in the loop” is not necessarily accurate given that humans tend to be biased towards accepting the recommendations made by computer programmes.175