Research methods often define a clear standard process of conducting research. These methods have evolved over decades of use, debate and even conflicts within academic communities (Alastalo, 2008). They allow textbook writers to give even `cookbook' kinds of recommendations on how to answer a particular research question. Therefore, anyone applying the methods must focus on following the presented process and conducting the steps as clearly and well as they can. The research product is then compared against these `established' practices to ensure it follows the practices and approaches of the community.
With many computational methods, these discussions are still emerging and responding to new kinds of data and approaches. For example, while classics in social network analysis are several decades old now, it remains unclear how to conceptualise `friendship' in online environments, where it appears to be looser than in traditional studies. In algorithmic data analysis, we are lacking decades-old traditions within social sciences. Disciplines like machine learning and computational linguistics have acknowledged some of these challenges (e.g. Wallach et al., 2009a; Chang et al., 2009; Schofield and Mimno, 2016; Wallach et al., 2009b). However, when working with social sciences, these findings need to be both translated and applied to conceptual and theoretical frameworks used. Social scientists are continuing basic research on validity and reliability of these methods within the social sciences and proposing research processes that would acknowledge these (e.g. Burscher et al., 2015; Wilkerson and Casas, 2017; Nelson, 2017; Denny and Spirling, 2018). Similarly, the constructive research methods we discussed are emerging in the toolbox for social sciences; the methodological reflections are emerging while we are using these. Furthermore, computational analysis often includes several steps, each of which has its own decisions about operationalisations, parameters and other factors that impact the outcomes.
Therefore, if I need to choose how to apply a specific method to a research problem, at the moment I cannot search for a textbook approach. There is no overall `cookbookâ of computational social sciences. Instead, I would search for research articles that have examined a similar research problem in the discipline or similar disciplines. Then I would imagine how the same approach might appear in my research context. What is the difference between the example articles and my research questions? When reading more technical papers, this gap is naturally much wider; when reading papers that examine the same research question but with different data, the gap is narrower. Often the case is somewhere between these two extremes. Then I would adopt the approach proposed in those papers into my own research. However, when closely examining various papers, it may be that different authors have different choices about the step-by-step process or some details are omitted in research reporting, which is surprisingly common in all sciences. Therefore, my version of the research process might be different from other scholars who think about the process as well. (This is also why I have avoided providing definitive guidelines in this book and sought to articulate at a high and generic level how specific methods presented could be applied.)