Just AI: Using Socio-Legal Studies of Fairness to Inform Ethical AI in Government

Artificial Intelligence (AI) is rapidly being deployed in government with potential for both enhanced service delivery and policymaking, as well as producing errors and human harms at pace and scale. Understanding and responding to the socio-legal implications of AI in government decision-making is crucial for citizen rights and accountable government. Current research into ethical and responsible AI has focused on measuring accuracy and bias, identifying ethical AI principles and creating legal and regulatory frameworks. Too often, however, the concepts and empirical work has been disconnected and ignorant of past research. While AI is new, the automation of government is not. This chapter builds on Adler’s pioneering work starting over 50 years ago to investigate how computerisation may support human wellbeing through fair procedural processes. It begins by reviewing Adler’s application of models of administrative fairness to charting the changes in social security systems from the late 1980s to the early 2000s. Adler’s insights about computerisation and procedural justice are then discussed in dialogue with contemporary ethical AI research. The chapter argues that there is an urgent need to draw deeper on long-established knowledge, to focus less on tools and more on designing systems with AI to advance wellbeing, to conceptualise the ethics of AI as contextual rather than solely technical, to distinguish between bias and unfairness and to undertake more rich-detailed ethnographic research. Just (or ethical) AI starts not with tools, but with co-designing socio-technological systems in multidisciplinary multi-stakeholder teams.