Work

Recurrent AI Assemblage Forms in Radiological Interpretation Routines

Type: article
Abstract: AI systems are expected to improve the efficiency and reliability of radiological interpretation. Prior research has examined cognitive engagement with algorithmic outputs or treated AI reading modes as stable workflow configurations. Yet we know less about how recurrent forms of AI integration are produced within the same routine. We address this gap through a qualitative study of an AI-based pulmonary nodule detection tool used in a cancer center. Contrary to expectations of stabilized integration, we find that the same artifact is recurrently assembled in three forms: sustained as a second-look verification device, desynchronized when material or clinical misalignments prevent actionability, and aligned when available as a first-look triage filter. We document a case in which these ostensively recognized forms are navigated performatively across examinations. These findings contribute to research on AI-supported organizational routines by explaining how an artifact can be recurrently enacted in multiple forms within the same routine.
Language: English
Open Access status: green
OpenAlex ID: W7222992577
Cites: 29
Cited by:
Related to: 10