Public data, private collaborator: Will machine learning relocate medical knowledge? (360G-Wellcome-213552_Z_18_Z)

This study focuses on the transformations in the creation and ownership of medical knowledge that result from applications of machine learning (ML). In medical image recognition, ML applications are currently being developed and tested to assist in diagnostics. These developments are often carried out as collaborations between public and private sectors, with public medical institutions providing data and medical domain knowledge, and private technology companies providing ML expertise. However, the algorithms implemented in these tools are typically proprietary, trade secrets that underlie the competitive advantage of the companies that develop and operate them. There consequently limited transparency into and access to the medical knowledge that they generate. This knowledge is thus privatised in the sense that it is encapsulated within closed software that forms part of the IPR of a private company. This study analyses such risks of privatisation and the role of open data in medical research. By comparing cases of ML applications in Ophthalmology, by conducting semi-structured interviews with stakeholders and organising two participatory workshops, the project will examine the role of data ownership and explore measures that allow medical knowledge to be kept in the public realm while still attracting private collaborations for ML applications.