29 September 2020

Government AI

'The use of artificial intelligence by government: parliamentary and legal issues' (NSW Parliamentary Research Service, 2020) by Daniel Montoya and Alice Rummery comments 

Digital technologies1 such as artificial intelligence (AI) are changing all human activities, at a pace that many commentators argue is increasing. There is currently no internationally recognised technical or legal definition of AI. Commonly misconceived as being a “single thing”, AI is better understood as a collection of general-purpose, “advanced digital technologies that enable machines to reproduce or surpass abilities that would require intelligence if humans were to perform them”. 

This paper focuses on the parliamentary and legal implications of governments using a form of AI: automated-decision making (ADM), which is deployed in automated decision-making systems (ADMS). While the paper reviewed a growing body of literature on the legal implications of ADM, and the broader range of ethical, social and political opportunities and challenges, little appears to have been published on the implications for Parliament’s capacity to exercise its scrutiny and lawmaking functions. Parliamentary responses to the rapid adoption of ADMS by governments around the world are in a similar phase, with existing legal and regulatory frameworks not always being fit for purpose. 

ADM promises many benefits, but also holds significant risks. The same qualities of ADM that may improve efficiency, timeliness and fairness in the public sector could also produce wide-scale negative outcomes for large numbers of people. Leading scholars have also observed that the tensions between automation and the foundational values of public law are likely to escalate with increased adoption of ADM. 

The NSW Government has been using ADM for over 20 years: for example, in November 1999 the Department of Fair Trading launched a business-name registration process that included a “substantial level of ADM”. The Commonwealth has used ADM since 1994, and currently automates hundreds of millions of administrative decisions every year. Recent developments demonstrate the increasing salience of this topic. In NSW, mobile phone detection cameras using AI have enforced illegal use of mobile phones since 1 March 2020, and the Government is due to release its NSW AI Strategy in the near future. In November 2019, the Commonwealth Government agreed with orders by the Federal Court of Australia that Centrelink’s online compliance intervention (OCI) system (‘robo-debt’) was unlawful. Globally, debates concerning surveillance technologies such as facial recognition have taken on new significance in light of proposed and existing government responses to the COVID-19 pandemic. 

After defining AI and ADM and providing some public sector examples, this paper discusses key parliamentary and legal issues in-depth. The paper finishes with case studies of selected parliamentary developments and recommendations from the literature on how Parliaments could respond.