Participation, prediction, and publicity: avoiding the pitfalls of applying Rawlsian ethics to AI

Abstract Given the popularity of John Rawls’ theory justice as fairness an ethical framework in artificial intelligence (AI) field, this article examines how fits with three different conceptual applications AI technology. First, discusses a proposition by Ashrafian to let agent perform deliberation that produces Rawlsian social contract governing humans. The discussion demonstrates inviability such application it contradicts foundational aspects theories. An exploration more viable context follows, introducing distinction between intrinsic and extrinsic theoretical adherence, i.e., difference approaches integrating system design those situating systems Rawls-consistent policy/legislative frameworks. uses emerging legislation EU U.S. well Gabriel’s argument for adopting publicity criterion field examples adherence theory. A epistemological challenges predictive then illustrates some implications While can make short-term predictions about human behavior fairness, long-term, large-scale results do not adhere theory, but instead constitute type utilitarianism Rawls vehemently opposed. concludes overview these arguments policymakers regulators.