Predictive algorithms in healthcare: constituting ‘Artificial Intelligence’ (AI) as near human

Abstract This article examines the development and integration of predictive artificial intelligence (AI) in clinical cardiology in Denmark. Employing a conceptual lens of nearness , I analyze how researchers and cardiologists unsettle and redraw boundaries between human and artificial intelligence. Based on ethnographic fieldwork on the CARDIA IHD algorithm, which predicts survival prognoses for patients hospitalized with ischemic heart disease, I demonstrate how AI is alternately enacted as a near-human ‘wingman’ or ‘butler’ and as an inferior, subhuman tool. While researchers rhetorically position the algorithm as a potential and valuable substitute to human reasoning, in clinical practice, its sometimes clinically unintelligible predictions lead cardiologists to disengage from it and exclude it from their decision-making. I argue that, for algorithms to acquire near-human qualities in practice, they depend on human hosts who experience affective-moral obligations and who are called to substitute for and care for the inadequacies of ‘artificial’ intelligence. The paper advances nearness as an analytical framework for examining transformations in how the category of the human is understood, experienced, and enacted in biomedical research and clinical care, particularly in contexts promoted to entangle human and ‘artificial’ intelligence.