The Value of Interdisciplinarity for Responsible Data Science
Interdisciplinarity: it’s a common enough word in academic settings. It’s a goal in many institutions to blend approaches, to utilize different worldviews effectively in addressing the big issues.
Interdisciplinarity is a normative concept; one that strives to value the space between disciplines or moving beyond strict disciplinary thinking in order to conduct research. There have been over the decades many discussions of whether the word should be “multi” or “inter” disciplinary, as a primary challenge to blending disciplines is, in fact, the institutionalized separation and distinction of approaches. And as the concept gets used more and more in everyday policy documents or reports of how research might be improved, it certainly gets more watered down and less clear. While we –as researchers in academic settings– might loosely agree that the goal of interdisciplinarity is important, many of us would be hard pressed to articulate clearly what it means, how multi- is different from inter-disciplinarity, and what it involves on a daily basis.
This white paper and accompanying seminar series is born out of a joint discussion of how data science, social sciences, and humanities can come into better alignment, not necessarily to lose the strength of individual disciplinary approaches, but to more strongly value, comprehend, and then utilize these strengths by seeing how they might connect around a common problem. Taking the current topic of misinformation campaigns, the challenge is to consider how a team can use the best of two seemingly different worlds (humanities and computing sciences) for tackling this troubling and persistent problem of the viral spread and uptake of misinformation.
We might move forward with basic questions such as: How much does a scientist’s training in a specific discipline affect how they see and study the world? The simple answer is “A lot.” A more complicated conversation arises when we ask further: How much can scientists benefit from the knowledge domains outside their specialisation, to better understand and solve critical issues of concern in the world? Again, we might answer, “a lot.” But this is a simplification. Rather, we ask: How can we learn the limits of our own ways of knowing? What habits of practice or vocabularies are built into disciplines and how can our viewpoints be broadened without losing the strength of disciplinary knowing? Is interdisciplinarity best practiced by learning and incorporating other epistemologies and methods or rather, is it less change-oriented and more value oriented? In other words, is it more a matter of appreciating these other ways of knowing and learning the vocabularies enough to value them in relation or comparison to our own disciplinary perspectives?
These are open questions, persisting over decades of discussions about multi-disciplinarity and inter-disciplinarity. We also recognize the many challenges to practicing anything we might call interdisciplinarity.
The first challenge is that contrasting perspectives are not easily reconciled. Even the most fundamental vocabularies in one’s discipline may be misunderstood or defined differently in other disciplines. In our case, take for example the notion of “people.” In the social sciences, the human is most often the central object of the analytical gaze and therefore a concern from the outset. Humans are what one studies. In computing sciences, humans are often only considered relevant toward the end of a study or project, when the outcomes of one’s software need to be tested, or are deployed for use by people. One’s research may be tested or relevant for “end users.” Such a contrasting operationalization of people in the research design is remarkable once we start to focus on it more closely. But these differences often remain invisible: they are such fundamental assumptions in research design, we don’t realize others don’t share these assumptions.
The second challenge is that interdisciplinarity is harder than disciplinarity for most researchers. It is comfortable to stay in one’s “home” discipline because it is familiar terrain. Once we develop a particular skillset, and hone this over time, participating actively –that is, conducting research in external disciplinary domains require us to learn new skills and get out of our comfort zone. If the paradigmatic assumptions are extremely different, this may require unlearning and relearning, both of which are positions of vulnerability especially in knowledge production environments, where expertise is the currency of trade.
The third challenge is that interdisciplinarity takes longer than disciplinarity, in all phases of research design, enactment, and reporting. Classic group decision making theories tell us that even though decisions made by groups are generally stronger and more accurate than those made by individuals, these processes take longer. This is primarily because different perspectives need to be heard, definitional assumptions and basic premises about how the world works must be articulated and further explicated until all participants have a basic comprehension, and then debates about how the process should be proceed must be agreed upon. These are often steps in decision-making that can be skipped when all participants have common baseline assumptions or have worked in similar circumstances. But when members of a research team are unfamiliar with the practices of science of their collaborators, this co-learning will take time. The payoff for such a time commitment is often weighed against the efficiency of projects performed by practitioners in the same discipline.
We value these challenges and, rather than dismissing them in our joint projects moving forward, have committed to a series of conversations to build common understanding. As this positively impacted our own project, we sought to expand this conversation by developing a series of conversations in 2022 with participants from multiple, disparate disciplines, across countries and levels of expertise. The outcome is a Responsible Data Science Seminar Series, co-sponsored by the Digital Ethnography Research Centre at RMIT University in Australia and The Minderoo Center for Technology and Democracy at University of Cambridge, UK. Six sessions in 2022 address various subtopics around the ethical value and logistical challenges of interdisciplinarity in building toward more responsible data science. These form useful heuristics for hosting conversations across multi-stakeholder groups and open possibilities for greater appreciation for the value of, as well as challenges of, interdisciplinarity. The agenda for each follows, and at the conclusion, we will be publishing a toolkit or best practice guide.
- Starting the conversation: Core values and challenges for ethical approaches to automation in society
- Bringing more diverse voices to the concepts of data, science, knowledge, and interdisciplinarity
- Logics (Reconciling different systems for inquiry to build better frameworks for collaboration)
- Layered Frameworks (combining efforts and models without losing discipline-specific definitions and traditions)
- Large and Small Scales (Focusing on issues and topics of critical global importance from bottom up as well as top down and data driven approaches, combining rather than privileging only certain scales)
- An agenda for value propositions and practical guides for interdisciplinary and responsible data science
15 May, 2022, Melbourne,
Annette Markham, Jenny Zhang, and Jey Han Lau
Bhaskar R, Frank C, Høyer KG, Næss P, and Parker, J. Interdisciplinarity and Climate Change: Transforming knowledge and practice for our global future. 2010. Routledge.
Chettiparamb A. Interdisciplinarity: A review. The Interdisciplinary Teaching and Learning Group. 2007. University of Southampton, UK.
Kerr NL, Tindale RS. Group performance and decision making. Annual Review of Psychology. 2004; 55:623–55.
Levine JM, Moreland RL. Progress in small group research. Annul Review of Psychology. 1990; 41:585–634.
Neff G, Markham A, Zhang X, Lau JH. Core values and challenges for ethical approaches to automation in society. Responsible Data Science Seminar Series, 1. 2022. Unpublished transcript.