From Prediction to Horizon: Clinicians' Negotiations of AI-Driven Personal Prognoses in Clinical Practice.

Abstract

Computational methods and tools under the label 'artificial intelligence' (AI) are increasingly promoted as solutions to the challenges of under-resourced and understaffed healthcare systems, with predictive modelling positioned as a means to improve efficiency and individualise care. Yet little is known about how predictive algorithms are taken up in the everyday practices of clinical decision-making. Drawing on ethnographic fieldwork with Danish cardiologists working with the CARDIA^IHD algorithm, this article examines how AI-driven personal prognoses intersect with clinical reasoning and organisational temporalities. We show, first, that complex algorithms introduce new forms of interpretive work; second, that their mathematical non-linear reasoning often conflicts with clinicians' causal and chronological frameworks; and third, that actionable horizons do not simply emerge with a survival prognosis but must be situated within specific times, places and organisational demands. Conceptualising prognostication with AI as horizoning, we argue that algorithmic personal prognoses do not straightforwardly deliver actionable futures, but must be anchored within the temporal and organisational conditions of clinical work to guide decision-making. Our findings advance sociological debates on the future by demonstrating that in algorithmically informed healthcare, the future is not a pre-given horizon but a negotiated, relational and socially enacted terrain.