Sustaining the practitioners’ tacit knowledge in the age of Artificial Intelligence: new challenges emerging through multiple case study[La pérennisation du savoir tacite des acteurs métier à l’ère de l’Intelligence Artificielle : émergence d’enjeux inédits à travers une étude de cas multiples]

This paper explores the perpetuation of practitioners' tacit knowledge in the context of projects aimed at designing Artificial Intelligence (AI) uses in organizations. By comparing an interdisciplinary review of the state of the art on tacit knowledge with an observational field study of 7 application cases in France and Switzerland, this article sheds light on the dynamics of capturing practitioners' tacit knowledge during the design and operation of AI models and highlights three areas for consideration: (1) the emergence of new devices for translating practitioners' know-how into data models and capturing tacit knowledge through the maieutic carried out in the design phase, (2) the difficulty of taking unconscious tacit knowledge into account when judging AI in use, revealing issues of interpretability, cognitive bias and trust, and (3) the capture of knowledge, including tacit knowledge, as the primary goal of Data Science projects. But this capture may not be desired by the practitioners or even introduce an intermediation that prevents the development of further tacit knowledge derived from real-life experience in favour of that linked to the use of AI. These considerations lead to the improvement of tacit knowledge perpetuation devices, as long as their legitimacy is justified, and the risks are mitigated.