Answers are grounded in this paper and cite their sources.
Governments increasingly rely on machine learning algorithms to make decisions, but the opacity of these systems impedes citizens’ ability to scrutinize state power and undermines democratic accountability. This paper evaluates two prominent approaches to explaining algorithmic systems — counterfactuals and transparency — by focusing on how they change the power dynamics between AI experts, government officials, and the public. I argue that both create problematic relationships of dependence despite their promise of empowering individuals. I propose a different approach to explanation that aims to facilitate public scrutiny of the power exercised by algorithmic systems and assign responsibility for the way they distribute benefits and burdens. This requires information that is intelligible to the public, normative, and systemic. I argue that systemlevel justifications that appeal to politically determined standards would empower the public to contest algorithmic systems and hold those responsible for them accountable.
No takes yet. Share an insight, caveat, or question.
Zeynep Pamuk (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: