Moraliser les machines communicantes. Des barricades morales à l’éthique située : trois cas d’usage de l’IA en milieu professionnel

The lightning spread of ChatGPT has kicked up a notch the panic that debates on artificial intelligence, and in particular on the morality of their linguistic and communicative performances, have been stirring up for over half a century. In this article, we will develop this questioning in two stages. First, we propose a documentary anthropology of the major ethical framing texts on AI, focusing on its impacts on communication at work. After highlighting the specificities of the universalist approach of these texts, we examine three case studies of conversational agents and communicating machines in professional contexts, documented by interviews with users and developers. The aim is to show that ethnographic analysis of practices reveals situated ethics putting the principles of a universalist approach on machine communication into a different perspective. This leads, in conclusion, to a rethinking of the role of the researcher, less an "expert" enunciator of rules and principles, and more an analytical spokesperson for his or her field.