'End-to-end' ethical AI. Taking into account the social and natural environments of automation

Antonio A. Casilli (1, 2, 3)

1 NOS - Numérique, Organisation et Société
2 SES - Département Sciences Economiques et Sociales
3 DiPLab - Digital Platform Labor

End-to-end ethical AI requires a consideration of the conditions of production of the data, tools and equipment used to manufacture and market these systems. A number of consumer products already apply this kind of ethical reasoning. For example, footwear and processed food manufacturers do not discriminate against their consumers based on gender, location or other factors – they are not as biased as AI in this respect. However, that does not make them ethical. A company is considered ethical if it respects workers’ rights, provides decent working conditions and minimises its environmental impact. In a similar vein, an ethical AI must minimise negative externalities both within and outside human communities and the same with regard to its production processes. Taking account of the human, social, political and economic contexts of today’s AI technologies can provide novel insights and suggest directions for future action.

Domains: Sociology

Antonio A. Casilli. 'End-to-end' ethical AI. Taking into account the social and natural environments of automation. Aida Ponce del Castillo. Artificial intelligence, labour and society, ETUI, pp.83-92, 2024, 978-2-87452-707-4. ⟨hal-04543418⟩