Developing AI for Medicine: Human Exceptionalism in Practice

Rocet Lecture Series hosted by the Socio-Technical Studies group (SPAIS) & Bristol Digital Futures Institute

Al continues to be subject to controversy and debate, both in its operation and potential impact. We argue that a key element of the debate is human exceptionalism (HE) – the idea that humans possess abilities and characteristics quite distinct from those of objects, artefacts and technology. This paper critically assesses the concept of HE and its use in practice.

Drawing on an extended ethnographic study of Californian based practitioners developing Al for medicine, it is argued that Al development is best characterized as a "creative muddle". Many very different actors were involved in the development process. Although end products emerged, our observations have the overwhelming sense that the practitioners did not know what they were doing, that they were being asked to develop for a practice that did not exist and that they disagreed among themselves. Against the common perception of technology development as logical and goal driven, we were struck by the apparent impurity of the process. By analogy with STS studies of the relation between philosophical discussions about science and science in practice, we find that Al developers deploy a complex mix of philosophical perspectives. In particular, development practice features four main varieties of HE. HE is not so much a philosophical driver of practice as a discursive resource which is drawn upon to characterize, criticize, explain and justify moves in the development process.