Optimising cohort data in Europe

will inform data subjects both in terms of strategies (i.e. how participants are reached) and in terms of content (i.e. what aspects of the study and potential data processing are disclosed) (Verhenneman et al., 2020). However, in the context of data collection with digital technologies, such transparency is difficult to achieve. In practice, there are still major trade-offs between the accessibility of information (and its readability by data subjects) and its completeness and accuracy (Marelli et al., 2020). Hence, big data research in general and data collection throughdigital technologies in particular, cannot achieve transparency, at least not in the sense required by the GDPR and current data guidelines. This can be explained by the opacity inherent to most of the digital infrastructures for data processing. This opacity takes various forms and has various origins (e.g. proprietary regimes, inaccessibility of algorithmic codes, and lack of readability of machine learning outputs) but its impact on participants’ ability to consent results in a common pattern: participants are alienated from the processing, analysis and collection of their own data. Namely, digital communication technologies often produce secondary data, meaning that there is no direct contact between the researchers and the participants (Roddy and Robinson, 2021). Finally, the GDPR fails to balance between requirements for data minimisation on the one hand, and the data subjects’ rights on the other. Namely, per GDPR, anonymisation applies when there is no longer any scientific or statistical purpose for keeping data in a form where it can be linked to a specific person. However, the GDPR never clarifies when and why these purposes are no longer relevant (van Veen, 2018). 3. Best practices, solutions and implications for implementation To solve the ethical concern of autonomy and consent, the following has been suggested: y y Dynamic consent arrangements should be the preferred option. This approach will give participants the opportunity to respond to changes (with respect to what the participant initially consented to) in the course of the research caused by unforeseen events, and in particular to the perceived advantage of future data sharing or new data uses (Borry et al., 2018; McMahon and Denaxas, 2019). Dynamic consent meets participants’ expectations of ownership and control over their data (Manzoni et al., 2018). y y The conditions and modalities of consent should be carefully considered and identified in relation to the participants. In particular, researchers should use established strategies to ensure that participants fully understand the content of consent. y y Techniques of “supported decision-making” where more active counselling and education should be used to overcome participants’ informational inequality (Manrique de Lara and Peláez Ballestas, 2020).