When two science disciplines meet: Evaluating dynamics of conjunction. The encounter between astrophysics and artificial intelligence

Abstract

This article points out some issues raised by the encounter between astrophysics (AP) and a newly emergent mathematical tool/discipline, namely artificial intelligence (AI). We suggest that this encounter has interesting consequences in terms of science evaluation. Our discussion favors an intra science perspective, both on the institutional and cognitive side. This encounter between machine learning (ML) and astrophysics points to three different consequences. (1) As a transverse tool, a same ML algorithm can be used for a diversity of very different disciplines and questions. This ambition and analytic intellectual architecture frequently identify similarities among apparently differentiated fields. (2) The perimeter of the disciplines involved in a research can lead to many and novel ways of collaboration between scientists and to new ways of evaluation of their work. And (3), the impossibility for the human mind to understand the processes involved in ML work raises the question of the reliability of results.