Automation in social security: Implications for merits review?

Abstract

This paper reviews illustrative aspects of the introduction of artificial intelligence into Australian social security's (Centrelink) original decision making and merits review. The review highlights the complexity and nuanced character of evaluating greater reliance on artifical intellignce over human decision‐making systems. It argues that technological change is inevitable, and is not necessarily either an unadulterated boon or bane, but calls for careful planning and a comparative assessment of the benefits and limitations of artificial intelligence versus human systems. In social security, it is concluded that particular attention should be given to avoiding discrimination against citizens who are technology poor or otherwise vulnerable.

“Sharing Is Caring:” Australian Self-Trackers' Concepts and Practices of Personal Data Sharing and Privacy

“…This research could be extended by focusing attention on social groups in Australia who have been pushed or coerced into self-tracking, who are disadvantaged or socially vulnerable, who are generating health data that are potentially stigmatizing or those who have experienced privacy harms or discrimination from their personal data being accessed by third parties. Australia has a recent poor record in the misuse of personal digitized information for exacerbating socioeconomic disadvantage and exerting surveillance over already under-privileged groups ( 34 , 64 ). Among other groups, Indigenous people have called for better data sovereignty, involving self-determination of what information is generated about them and better control over third-party access to their data ( 65 ).…”

“…As is the case in other countries, Australians who generate personal information about their everyday activities using online or mobile technologies are open to potential misuse of their data ( 31 ). Furthermore, several highly publicized personal data breaches and cases of government misuse have occurred in Australia in recent years ( 33 , 34 ).…”

Introduction to the Special Issue on Big Data and Social Policy in Australia

“…Some Australians will be unable to answer the question of whether the Australian Government can be trusted to use linked data fairly without thinking of the disastrous Online Compliance Initiative for debt recovery – commonly known as Robodebt. Launched in 2016, this initiative involved the automatic generation of debt notices using a faulty algorithm that averaged income across the year and relied on the cross‐matching of Centrelink data with Australian Taxation Office records (Carney, 2020). While there is history dating back to 1991 of these records being cross‐matched to check for overpayment, the removal of the human from “the loop” was key to the ramping up of the debt recovery programme from the discovery of 20,000 overpayment in 2015–2016 to nearly 800,000 in 2016–2017 (Whiteford, 2020).…”

Technical considerations when implementing digital infrastructure for social policy

“…The incorporation of big data and advanced computational methods into social policy development and research is becoming increasingly common as we continue to experience rapid technological change (Carney 2019). Yet, despite the substantial body of research emerging on the characteristics and infrastructure challenges related to big data, integrating its use into social policy projects and research continues to present challenges to practitioners.…”

“…The exponential growth of big data in such a short period of time means there is limited research into how big data can actually improve government service (Hong et al 2019). In particular, how to weigh the potential benefits of big data in social policy and research against the significant technical and organisational resources required to safely and securely integrate it remains an open question (Carney 2019).…”