Moving further from a single research question, we move to discuss larger research projects which may have several research questions. It is as important to identify how computational sciences are brought into a larger research project, and ask questions such as: how are responsibilities defined and shared, how collaboration is nurtured, and, who can make interpretations about data? Even research on commercial applications of data science is just beginning to grasp this topic and making suggestions on how to organize data science projects (see for example Saltz and Shamshurin, 2016; Saltz et al., 2017). Like many other aspects of computational social sciences, there is not yet best practices into these questions. A common understanding is that both social sciences and computational skills are required to excel in research projects (Wallach, 2018; Grimmer, 2015). This suggests a need to create multidisciplinary teams to address different sides of the process, and is often proposed as a solution for this work. However, multidisciplinary is not as easy as one might hope. While working on a research grant asking us to explicate the methodology for multidisciplinary collaboration, the room nodded when one team member said: “this is a difficult bit to write because we are so experienced in multidisciplinary work and know it is not as easy as it sounds.”
Many might think that a collaboration between computational (“methodology”) experts and domain experts in a research project may be a good direction to run computational social science projects. The idealized version of such collaboration is that both bring in their respective expertise and new ground-breaking research emerges from this. However, major challenge in multidisciplinary ventures like this is the lack of common jargon and concepts across the communities. Trading zones which help people from different communities to exchange their ideas are essential with this (Galison, 1999). There can be different kinds of trading zones, some more collaborative, others use power to coerce collaborations. Similarly, if the aim of the collaboration creates a new homogenous culture or two enhanced heterogenous cultures are different types of trading zones (Collins et al., 2007). To articulate these in computational social sciences McFarland et al. (2016) examined research outputs, or papers. In their view, there are three different shapes of computational social sciences: interdisciplinary joining where the outcome is a multi-lingual mosaic (showing more heterogenous culture), the interdisciplinary embedding where the secondary discipline seeks to support work in the primary discipline (which can be collaborative or cultures), and third - and most challenging - full mixing between these disciplines (driving for homogeneous culture). As this categorization shows, there are many ways to engage in interdisciplinary work which may lead to different outcomes. It is good to know about these opportunities and even discuss these expectations. In all of these cases it is essential to support how to support the collaborative work.
Collaboration often requires facilitation: supporting the collaborative work. Facilitation focuses on two steps: helping people to be included and sharing their understanding through participation, and formulating the participation into concrete outcomes through reification. Thus, in any trading zone, people should be allowed to engage in knowledge sharing, for example, through discussing the theme and addressing challenges. However, it is also important to materialise these somehow: in traditional facilitation workshop, mapping discussion topics as Postits and organizing them is an act of reification. In academic research, this reification can take place in several forms: examples of the data, which help the process forward, computed results which allow further investigation, or even mockups to help different participants to discuss about the proposed methods and outcomes from them. Mockups may be helpful if writing the code and running the analysis takes longer than expected. Thus, facilitating collaborative work requires a bit of thinking how to help inclusion and concretization.
However, this book really advocates a different approach to this. It seeks to challenge the assumption that methodology experts and domain experts should be different individuals or teams, and has (hopefully) developed basic skills to design and do data analysis with any of the methods illustrated in the book. My own experience is that developing such interdisciplinary research teams or organizations is difficult. I've been involved in many attempts to establish such research teams, and it has been always difficult to get a good idea of shared concepts or discuss correctly across the “methodology” and “domain expert” boundary. For example, our paper focused on the meanings of racism in Finnish media (Pantti et al., 2019) begun in during a long meeting where communication scholars explained that we should filter racists content from online discussions and me (as a methodology expert - with no previous background to racism as an academic research subject) explaining that to achieve this, we should be able to somehow define racism to filter such content. However, they said that the aim of the work is not to define what racism is but to study how it is present in the discussion. Luckily my colleague was able to rearticulate my concern and helping us to move forward with the discussion: we did not focus on racists discussions in the media but rather discussions on racism - much easier to focus on in computationally. Furthermore, I am not alone with these experiences. For example, () speaks about chasm of computers, highlighting how integrating computing into biology required to some degree computer scientists to learn biology as well.
I believe that high-quality scholarship often requires both novel insights in social sciences and novel insights in opportunities allowed by technology. Or said differently: both sociological imagination and technology imaginations are required to identify interesting research questions and articulations against established scholarship and ability to transform those ideas into computational code. And emerging those innovations may require the ability to integrate ideas, findings and concepts across these disciplines in novel ways. Such work often requires people who have a good understanding of both disciplines to ensure that ideas are not lost in translation or trivialized. None the less, even in these cases those scholars must focus on how to navigate academic fields successfully.
Finally, people come in the multidisciplinary settings with different backgrounds. van Wijk (2006) suggests that this creates two gaps which may hamper collaboration: knowledge gaps and interest gaps. Above we mostly examined knowledge gaps: to collaborate people must create a trading zone and shared language to fill in the knowledge gaps. He also argues that there is an interest gap in interdisciplinary collaboration. There are different institutional expectations about what the research should produce. For example, for machine learning community, the expected outcome is often an improved methodology which can be used across various domains. For social scientists, the goal is to describe what is going on in the society. The former requires fancy and novel methods, whereas the latter may be even hampered by such approaches. Novel methods require may not follow the customs and traditions of that the field, thus challenging if they can be used to answer an empirical question. These two gaps can create tensions in any collaboration relationship as participants may have different incentives about doing research. Incentive structures are important to consider as they relate to participants career advancement and thus should be considered in any project work.