Administrative Law and the Governance of Automated Decision-Making: A Critical Look at Canada’s Directive on Automated Decision-Making
Forthcoming, (2021) 54:1 University of British Columbia Law Review
29 Pages Posted: 10 Dec 2020
Date Written: October 30, 2020
Abstract
The adoption and use of automated decision-making in government is a growing trend that raises concerns about fairness, transparency and accountability. An emerging issue is how to govern automated decision-making to ensure fairness and accountability. Most research in this area has focused on privacy, data protection and transparency issues. However, long-standing principles of administrative law have shaped the adoption of administrative decision-making processes and guided oversight of their fairness and accountability, and administrative law can provide an important lens through which to assess the use of automated decision-making by governments.
In 2019 Canada's federal government adopted a Directive on Automated Decision-Making (DADM) and an accompanying algorithmic impact assessment (AIA) tool. The DADM and the AIA tool offer a framework for the adoption and implementation of algorithmic decision-making by governments. The influence of common law administrative law principles is evident in the DADM. This paper assesses the DADM through an administrative law lens, and asks to what extent the Directive meets, exceeds, or falls short of principles of administrative fairness. In doing so, it also considers whether the reliance on long-established administrative law principles to guide the adoption of automated decision-making processes may create blind spots in which fundamental differences in the nature of the decision-making process leave those subject to automated decision-making with an inadequate level of transparency and due process. Ultimately, the paper explores the role and shape of administrative law principles in an era of automated decision-making.
Keywords: artificial intelligence, automated decision making, AI, administrative law, algorithmic impact assessment
Suggested Citation: Suggested Citation