Abstract

A normative framework grounded in Kantian ethics that provides a practical decision procedure for evaluating AI systems. The Kantian Filter introduces three tests universalisability, human dignity, and non-delegable responsibility - to address the responsibility gap in algorithmic decision-making.