File(s) under permanent embargo
Oaths and the ethics of automated data: limits to porting the Hippocratic oath from medicine to data science
journal contributionposted on 2022-02-23, 00:00 authored by Kate MannellKate Mannell, R Fordyce, S Jethani
This paper argues that the proposal for a ‘Hippocratic oath for data science’ is a severely limited form of data ethics for automated culture. Drawing on the oath used within medical professionalism, proponents as diverse as Wired and the European Data Protection Supervisor have argued for a Hippocratic oath for data science as a way of introducing a soft regulatory environment. In this paper, we analyse the history of the Hippocratic oath and the professions of medicine and data science to suggest that this proposal offers an individualized solution to systemic problems and, as such, is unlikely to be effective. We further argue that the proposal of the Hippocratic oath ignores the degree to which the profession of the physician is different from the profession of the data scientist in ways that limit the transfer of an ethical framework between them. In particular, we note that automated data access leads to a lack of clear professional identity among those who act as data stewards which, unlike in a medical context, makes it unclear how breaches of an oath would be adequately sanctioned. We also argue that, unlike in a medical context, harms can be difficult to define and have historically been poorly acknowledged, making it difficult to meaningfully take an oath to ‘do no harm’. We propose that in the context of data science, a Hippocratic oath would provide little substantial protection for users and largely penalize workers over companies while deferring responsibility away from those profiting from data extraction. The paper concludes by suggesting that the limits of the Hippocratic oath are significant to the point that other regulations should also be sought, although proposals for oaths have value as catalysts for cultural change within the technology industry.