,
A Crack in the Algorithm’s Facade. A Fundamental Rights Perspective on “Efficiency” and “Neutrality” Narratives of Algorithms
(Dis)Obedience in Digital Societies. Perspectives on the Power of Algorithms and Data, transcript Verlag, Bielefeld, (марта 2022)
DOI: 10.14361/9783839457634-009
Аннотация
Victoria Guijarro Santos argues that the current conception of algorithmic bias as merely statistical inaccuracies is misleading. Deploying data-based algorithms to automate decision-making breaches fundamental rights particularly if these algorithms are not statistically biased but produce statistically accurate outputs. Guijarro Santos coins the inherent harms of data-based algorithms as the »generalization effect« and the »effect of cementing social biases«, amounting to breaches of fundamental rights guaranteeing autonomy, equality, and non-discrimination. Efficiency goals alone cannot justify these harms and technical solutions cannot remedy them. Data-based algorithms should not be deployed to automate decisions. Rather, they should be used for what they excel at: recognizing social patterns of oppression.
