Aarhus University Seal / Aarhus Universitets segl

Politics of automation

Risks and accountability in the domain of algorithmic selection

Florian Saurwein, Austrian Academy of Sciences and Alpen-Adria-Universität, Austria

Recent developments in economy and society are subject to far-reaching changes that are shaped by digitization, mobile communication, big data, robotics and artificial intelligence. Among these technology trends, one phenomenon of increasing relevance is „algorithmic selection” which builds the technological fundament for several Internet applications such as search engines, news-aggregators, recommender systems, scoring services as well as monitoring and forecast applications. The social relevance of algorithmic selection is caused by its wide and rapid diffusion, its increasing influence in a variety of social domains and economic sectors and the risks that accompany algorithmic selection, e.g., manipulation, bias and censorship. The rise of algorithms, their increasing influence and the involved risks trigger the debate about appropriate governance and accountability mechanisms. The presentation provides an overview on governance and accountability structures in the domain of algorithms which points to many actors who are involved in the governance of algorithms. From a governance perspective, the paper argues, that accountability structures are determined by application structures and the structures of control of algorithms. The variety of involved parties, the distribution of action and the broad spectrum of accountability conceptions point to a structure of “distributed accountability” in accountability networks. This involves the danger of a confusion of accountability and shirking of responsibility in the domain of algorithms.