An Algorithm at the Table at Multidisciplinary Diagnostic Meetings: A Dramaturgical Analysis.

Abstract

This paper focuses on how a radiological classification algorithm temporarily became part of existing diagnostic deliberations between different medical specialists. It describes a participatory ethnographic case study in which the output of the algorithm was presented by the first author during weekly multidisciplinary neuro-oncology meetings. Following the dramaturgical framework of deliberations by Hajer and adhering to the sociology of diagnosis, we focus on how the algorithm interacted with the established deliberative dynamics of the meetings. We show that by the algorithm's presence, physicians were prompted to speculate about what the algorithm could mean for their collective diagnosis. They did not foresee an independent or leading role for the algorithm yet speculated about how, when and by whom the algorithmic output should be presented. Some physicians followed dominant AI tropes of self-evidence and objectivity, whereas others sought ways to contextualise and question the algorithmic output. We argue that following a dramaturgical framework unveils important questions that must be addressed to make algorithms fit into existing diagnostic practices, such as who represents the algorithm, when and how.