Titel
More than the Sum of its Parts: Susceptibility to Algorithmic Disadvantage as a Conceptual Framework
Autor*in
Abstract
Algorithmic systems are increasingly being applied in contexts of state action to, in some capacity, mediate the relations between state and individual. Disadvantageous effects, such as potential discriminatory outcomes brought forth by different kinds of biases, have been the locus of severe critique by academic scholarship and political activism. There has been scholarly work conceptualizing biases and types of biases, as well as types of harm. Drawing from Elizabeth Anderson’s conceptualization of relational equality, this paper emphasizes the relationality of the encounters between state and individual. This paper introduces "susceptibility to algorithmic disadvantage" as a conceptual framework to address the relational constellation at play. Susceptibility to algorithmic disadvantage has a vertical dimension that addresses the relation between a state actor and an individual and a horizontal dimension that is characterized by intersectional inequalities that prevail in societal contexts. Intersectional feminist scholarship has established that interlocking systems of oppression amount to more than the sum of their single-axis parts. Paralleling this argument, this paper argues that susceptibility to algorithmic disadvantage amounts to more than the sum of the vertical and the horizontal dimension: the dimensions co-constitute and reinforce each other. The proposed framework is applied to four international case studies situated in crucial areas of state action: facial recognition in law enforcement in the USA, biometric identification in social welfare in India, dialect recognition in the asylum system in Germany, and grade prediction in the education system in the UK. Viewed through the lens of the proposed framework, heterogeneous use cases in different locations and areas of state action emerge as similar considering the inquiry into questions of justice, rendering the proposed framework a useful tool for analysis.
Objekt-Typ
Sprache
Englisch [eng]
Persistent identifier
Enthalten in
Titel
FAccT '24
Proceedings of the 2024 ACM Conference on Fairness, Accountability, and Transparency
ISBN
979-8-4007-0450-5
Verlag
ACM , 2024
Seitenanfang
909
Seitenende
919
Zugänglichkeit
Rechte
Rechteangabe
© 2024 Copyright held by the owner/author(s)
Universität Wien | Universitätsring 1 | 1010 Wien | T
+43-1-4277-0