Re-thinking Bias in the Age of Automation

Abstract

This article explores the applicability of the rule against bias to administrative decisions made wholly by automated systems, questioning the ability of Australia's administrative law system to adapt to increasing digitalisation and automation of government processes. By examining the behaviour and characteristics of neural networks, this article demonstrates the ways in which bias may infuse the decisions of predictive automated systems. It subsequently reveals the difficulties courts will likely face in transposing the rule against bias to automated decision-makers, thus advocating the use of alternative grounds of judicial review to remedy defective decisions. Ultimately, the relevancy and reasonableness grounds; despite their similarly human focus, are identified as the most amenable to judicial modification. Show more