The Inherent Politics of AI as a Matter of Justice

Abstract

Accounting for artificial intelligence (AI) related risks to social justice, the research field often called algorithmic fairness has developed. Within it, two subfields stand out: 1) formal algorithmic fairness, providing technocentric reforms, such as mathematical bias mitigation, and 2) structural approaches to algorithmic fairness, which emphasize societal reforms, addressing AI as a socio-technical structure and the structural injustices that shape AI. These two research paradigms of algorithmic fairness differ in the locus of intervention, the level of abstraction of AI, their scope, and their methods. I argue that their tendency to position themselves as seemingly incompatible oppositions creates a conceptual gap, so that neither formal nor structural approaches, alone or together, can provide a sufficiently comprehensive account to address AI-related injustices. Formal approaches tend to be too simplistic to account for structural complexities, whereas structural approaches tend to lose nuance regarding the specific characteristics of AI, thereby failing to explain whether and how AI-related injustices differ from structural injustices in society. To bridge this gap, I propose drawing on Winner's (1980) ‘inherent politics of technology’, according to which some technologies have political qualities as they require (or are strongly compatible with) specific political environments, defined by their internal or external characteristics. To be productive in the discourse on justice, I extend Winner's account by conceptualizing the inherent politics as 'justice-sensitive', meaning that the politics themselves are not inherently just or unjust, but become so in social contexts and are evaluated against specific theories of justice. While the exploitative external conditions that shape the development of AI gain more and more attention, the subtle political nature of AI’s internal technological characteristics is largely overlooked. To illustrate my account, this contribution will analyze predictive AI and its inherent, justice-sensitive politics that are caused by its internal technical characteristics as an example. Finally, I suggest that my outlined account of the inherent, justice-sensitive politics of AI caused by its internal technical characteristics provides a conceptual gain by identifying a new underexplored way in which AI contributes to (in)justice, as well as a practical gain by being able to devise better interventions based on the specific impact of the AI technique.