AI for pattern recognition and objectivity: the case of melanoma detection
Benoît Duchemann · 2022
This chapter focuses on the transformation of epistemic means of melanoma detection that have occurred since the beginning of the 1980’s: from the dermatologist’s experienced eye and cognition to the development of deterministic softwares using image processing to end with the presently final leap consisting of contemporary artificial intelligence aiming at a new form of objectivity. By questioning the models and the processes encapsulated within those digital technologies we show that this claim for an objectivity based on the exclusion of the human observer from the detection loop is still rooted in human subjectivity. Moreover, by considering the AI automata itself as an observation subject we show that although this new subject is to be competent it is selfcentred and cannot be considered as objective in a knowledge relation with a community of practice.