Automated Decision-Making Systems in Precision Medicine – The Right to Good Administration at Risk

Automated decision-making (ADM) systems – whose algorithms are based on Artificial Intelligence and more specifically on machine learning and deep learning – predict the likelihood of an outcome based on profiling the input data. ADM systems, which are predominantly developed by the private sector, are a promising device for the field of precision medicine, where medical intervention is based on the patient’s unique profile that consists of their genomic data, medical records data, environmental data, and lifestyle data. Such ADM systems are used when diagnosing or creating a treatment plan for patients. As these ADM systems take a bodily sample, for example from blood or human tissue, to predict which diagnosis or drug regime is most suitable, they are governed by the In Vitro Diagnostic Medical Devices Regulation in the European Union. However, the general features inherent to coding algorithms based on machine learning and deep learning – amongst others i) the self-learning ability, ii) the lack of a causal link, and iii) the use of intellectual property rights –, may form perils to the right to good administration that prescribes the legal norms of administrative conduct, including towards individuals. Particularly, the right to be heard, the right to access one’s file, and the duty to state reasons may face considerable hurdles. Thus, this contribution aims to scrutinise these risks to the right to good administration and proposes a research agenda to overcome them.