Quand l’intelligence artificielle théorisera les organisations

International audienceThis article explores the feasibility of machines inventing and theorizing organizations. Most machine learning models are automated statistical processes that barely achieve a formal induction. In that sense, most current learning models do not generate new theories, but, instead, recognize a pre-existing order of symbols, signs or data. Most human theories are embodied and incarnated: they spawn from an organic connection to the world, which theorists can hardly escape. This article is organized in three parts. First, we study the history of artificial intelligence, from its foundation in the 19th century to its recent evolution, to understand what an artificial intelligence would be able to do in terms of theorization... Which leads us, in a second step, to question the act of scientific production in order to identify what can be considered a human act, and what can be the subject of modelling and autonomous learning led by a machine. The objective here is to assess the feasibility of substituting man with machine to produce research. In a third and final part, we propose four modes of theoretical exploration that are already the work of machines, or that could see, in the future, a complete substitution of man by machine. We conclude this article by sharing several questions about the future of research in organizational theory, and its utility, human or machine, for organizations and society.Cet article 1 explore la possibilité qu’une intelligence machine puisse théoriser des organisations ; et qu’elle le fasse mieux qu’une intelligence humaine dans un proche futur. La plupart des modèles d’apprentissage des machines sont des processus statistiques automatisés qui sont à peine capables d’une induction formelle et ne génèrent pas de nouvelles théories, mais reconnaissent plutôt un ordre préexistant. Les théories humaines sont incarnées ; elles naissent d’un lien organique avec le monde, auquel les théoriciens ne peuvent échapper. Cet article envisage de surmonter cet obstacle pour accueillir une révolution théorique apportée par l’IA