Algorithmic responsibility without accountability: Understanding data-intensive algorithms and decisions in organisations

Abstract

Social science research has been concerned for several years with the issue of shifting responsibilities in organisations due to the increased use of data-intensive algorithms. Much of the research to date has focused on the question of who should be held accountable when ‘algorithmic decisions’ turn out to be discriminatory, erroneous or unfair. From a sociological perspective, it is striking that these debates do not make a clear distinction between responsibility and accountability. In our paper, we draw on this distinction as proposed by the German social systems theorist Niklas Luhmann. We use it to analyse the changes and continuities in organisations related to the use of data-intensive algorithms. We argue that algorithms absorb uncertainty in organisational decision-making and thus can indeed take responsibility but cannot be made accountable for errors. By using algorithms, responsibility is fragmented across people and technology, while assigning accountability becomes highly controversial. This creates new discrepancies between responsibility and accountability, which can be especially consequential for organisations' internal trust and innovation capacities.