Not Another Grocery List: Proposals for an Effective AI Ethics Implementation

Considerable effort has been invested during the last decade to develop ethical frameworks to guide the use of artificial intelligence (AI) technologies. Most of these frameworks take the form of an axiomatic list, that is, a list of principles or values taken as self-evident truths. This paper argues these lists face three challenges: divergence within the core principles, structural incoherence, and ethics washing. Three proposals are then put forward to address these challenges and ensure the effective implementation of AI ethics. First, an ethics framework must be parsimonious and, second, a framework building on multiple principles or values must also provide a balancing structure to deal with the different implications of these principles. The third proposal is to envision AI ethics as a process more than a static list of axiomatic principles or values.