Data generation methods across the empirical sciences: differences in the study phenomena’s accessibility and the processes of data encoding

Data generation methods differ across the empirical sciences. Today’s physicists and engineers primarily generate data with automated technologies. Behavioural, psychological and social scientists explore phenomena that are not technically accessible (e.g., attitudes, social beliefs) or only in limited ways (e.g., behaviours) and therefore generate data primarily with persons. But human abilities are involved in any data generation, even when technologies are used and developed. This article explores concepts and methods of data generation of different sciences from transdisciplinary and philosophy-of-science perspectives. It highlights that empirical data can reveal information about the phenomena under study only if relevant properties of these phenomena have been encoded systematically in the data. Metatheoretical concepts and methodological principles are elaborated that open up new perspectives on methods of data generation across the empirical sciences, highlighting commonalities and differences in two pivotal points: (1) in the accessibility that various kinds of phenomena have for the persons generating the data and for the researchers, and (2), as a consequence thereof, in the processes involved in the encoding of information from these phenomena in the signs (symbols) used as data. These concepts and principles cut across establish method categorisations (e.g., human-generated versus instrument-generated data; quantitative versus qualitative methods), highlighting fundamental issues equally important in all sciences as well as essential differences. They also provide novel lines of argumentation that substantiate psychologists’ and social scientists’ increasing criticism of their own disciplines’ focus on standardised assessment methods and establish connections to concepts of data generation developed in metrology.