Towards experimental implementations: Moving emerging artificial intelligence technologies into real-world clinical settings

Despite the many expectations and promises of artificial intelligence (AI) in healthcare, there is still a chasm between the claimed performance of complex algorithms and their deployment and scaling in contexts of care beyond research projects. This chapter examines the role of expectations in the early stage of AI deployment. The research is based on an empirical study of a competitive dialogue process initiated by a Norwegian hospital trust, aiming to procure commercially available AI applications for use areas within radiology and image diagnostics. The data, collected through non-participant observations of digital meetings and interviews, are analysed by drawing on perspectives and concepts from Science and Technology Studies (STS) that address the dynamics of expectations, technoscientific promises, and promissory work. The chapter explores both the characteristics of expectations and promises surrounding AI technologies for radiology and the role of these expectations as they are articulated by the vendors and selectively pursued by the procurement team. The study raises issues about how scholars within STS should respond to emerging technoscientific promises in contexts of care. It highlights that investigators—in order to avoid uncritically reproducing the rhetoric of the technology suppliers—need to scrutinise expectations as they enter the world of user organisations and are subjected to different kinds of assessments by the actors involved.