Computer science developed in an essential tension between a universal aim towards domain-free procedures and knowledge, and pragmatic and specific applications to fulfil administrative, commercial, industrial, and scientific needs. Considering science, I will address the social and epistemological relationships between domain specialists and computer scientists: this aspect is related to the old problem of the applicability of mathematics, but has been constantly renewed and extended in the second half of the 20th century through the growth of computer science. The current and fast development of AI and, consequently, AI-related institutions, policies, and funding, despite its assumed radical novelty, raises organisational and also epistemological issues similar to those experienced since the 1950s by scientists from various fields. This claim will be defended from an examination of the case of archaeology in France. Archaeology has three interests in this regard: 1) it has been shaped as a discipline during the same period as computer science; 2) it is part of the humanities, in which the use of quantification and automated computation was less evident to the practitioners than in the so-called “hard” sciences, so raising particular epistemological and social issues; however, 3) archaeology was one of the pioneering domains of the humanities and social science to develop computer application. Two questions will be addressed. First, the conditions of emergence of a speciality related to the use of quantitative methods or computers. Despite early applications of computing, and contrary to other fields such as ecology, biology, or neurology, there is no computer-based disciplinary speciality in archaeology (contra numerical ecology, biostatistics, neurocomputing, respectively). Second, the collective form of organisation to make computer scientists and domain specialists collaborate. Publications and archive documents were used to reconstruct the social space of these interdisciplinary relationships in France and its evolution. On the one hand, a general picture is given using a dataset with the participation of 1057 actors to 59 conferences, identifying three active groups of researchers. Archives and interviews were used for an in-depth analysis of these groups, illustrating three models of policies and socio-epistemic relationships between computers scientists and archaeologists.< Réduire