The Fall of an Algorithm: Characterizing the Dynamics Toward Abandonment

Abstract

As more algorithmic systems have come under scrutiny for their potential to\ninflict societal harms, an increasing number of organizations that hold power\nover harmful algorithms have chosen (or were required under the law) to abandon\nthem. While social movements and calls to abandon harmful algorithms have\nemerged across application domains, little academic attention has been paid to\nstudying abandonment as a means to mitigate algorithmic harms. In this paper,\nwe take a first step towards conceptualizing "algorithm abandonment" as an\norganization's decision to stop designing, developing, or using an algorithmic\nsystem due to its (potential) harms. We conduct a thematic analysis of\nreal-world cases of algorithm abandonment to characterize the dynamics leading\nto this outcome. Our analysis of 40 cases reveals that campaigns to abandon an\nalgorithm follow a common process of six iterative phases: discovery,\ndiagnosis, dissemination, dialogue, decision, and death, which we term the "6\nD's of abandonment". In addition, we highlight key factors that facilitate (or\nprohibit) abandonment, which include characteristics of both the technical and\nsocial systems that the algorithm is embedded within. We discuss implications\nfor several stakeholders, including proprietors and technologists who have the\npower to influence an algorithm's (dis)continued use, FAccT researchers, and\npolicymakers.\n

Ähnliche Arbeiten