October 1, 2021
Restricted Access , http://purl.org/eprint/accessRights/RestrictedAccess
Élise Berlinski, « The imaginary infrastructure of AI : an accounting technology », Theses.fr, ID : 10670/1.115366...
This thesis studies current developments in artificial intelligence (AI) and asks the following research questions: (1) Why is a great enthusiasm invested in these technologies even though they are poorly understood; (2) What grants power to quantifications, and specifically to AI quantifications? After proposing a conceptualization of accounting, we argue that quantifications become powerful when they become accounting. We study whether AI becomes an accounting technology and to do so we study the ideals associated with it. We show that they are the result of an imaginary as conceptualized by Cornelius Castoriadis in The Imaginary Institution of Society (1975, Éditions du Seuil, Paris). This imaginary is carried out by the action of the people, which participates symmetrically in giving it a shape, it is at the same time the result of socio-historical evolutions. We first trace the socio-historical-technical evolutions of AI before looking at the current context through three distinct empirical materials that allow us to link local and global levels. We argue that an accounting regime emerges with AI, which we characterize and propose that it helps understand changes in the accounting practice. We further argue that it is the specificity of AI quantifications, which are the result of practices and an imaginary, that allows us to understand the enthusiasm invested in them. Finally, we propose that the use of AI requires efforts of distinction so that it does not diminish the autonomy of people.