Optimising cohort data in Europe

al., 2018). A solution in this respect is to give back power to participants through participative structures and modalities inherent to communication devices (e.g. personalised log in, possibility to delete personal data) (Park and Skoric, 2017). y y Researchers are reluctant to involve vulnerable populations in an attempt to protect them from harm. As a result, vulnerable populations are not fully represented in cohort studies. Researchers should be aware of this and apply appropriatemeasures to be as inclusive as possible. y y Raw media datasets are often analysed according to distributed learning principles (i.e. artificial intelligence). This can reveal overarching patterns that could not have been visible with standard technologies, which in turn extends the scope and depth of data analysis (Merolli et al., 2013). However, raw data analytics can also reveal patterns that can be either biased or incorrect (Oltenau et al., 2019). Raw data sets are vulnerable to biases because their content is unstable from the onset (Taylor and Pagliari, 2018). Twitter data for instance, is produced by different actors (bots, humans and institutions) who each have different motives and incentives for their online interventions (Tene and Polenetsky, 2012). y y What counts is not only what is in the data (e.g. biases) but also what isn’t (e.g. excluded communities). Raw media data should thus be complemented by an in depth exploration about the nature of data and the conditions of its production. In any case, no policy should be based on the patterns derived from raw data alone (Nebeker et al., 2018). y y Data collection through communication devices may not be necessarily appropriate for certain communities. Namely, researchers have to demonstrate clarity about the type of research being conducted and the target population. In particular, researchers have to determine the extent to which the targeted communities are represented online (Martinez-Martin and Kreitmair, 2018). y y Communication technologies give the opportunity to reinterpret the issues of confidentiality and privacy in the collection and storage of media data. In the context of online discussion boards for instance, communication technologies generate a new type of research data that of “personal health text” (Bond et al., 2013). This requires a different way of handling data both on the practical and ethical level. y y On the ethical level, confidentiality and privacy are culturally and institutionally reproduced through the use of communication technologies. As a result, “traditional” perceptions of confidentiality may not always apply in online settings. For instance, participants with personal blogs and twitter accounts under their own names may not necessarily have anonymity concerns: they may instead want to be cited as much as possible (Keusch et al., 2019). y y The characteristics of communication technologies may help researchers to adapt confidentiality and privacy issues according to participants’ needs and online landscapes. For instance, complete anonymity can be enforced by muddling