Epistemic Injustice and Government Information Systems: Lessons from Two Cases

Joris Hulstijn, Huimon Dong, Réka Markovich

Abstract: Cases of bias and unfair decisions in automated decision-making are heavily discussed. When unfair decisions can be attributed to a difference in the knowledge of groups of subjects, we can speak of epistemic injustice (Fricker). In this paper, we analyse the various types of epistemic injustice: testimonial, hermeneutical, distributional, and content-focused epistemic injustice, and show how they can be conceptualised. We apply the notion of epistemic injustice to analyse what went wrong in two automated decision making scandals: Toeslagenaffaire (Netherlands), and RoboDebt (Australia). We discuss key observations from the cases and show that they can be categorised as various types of epistemic injustice. Based on these observations, we draw lessons about governance of automated decision making systems.