In 2009, the University of Amsterdam opened its new Faculty of Sciences building, which would house all of the natural sciences under one roof. When we moved in, one thing became immediately clear: they had forgotten the walls. Except those required for structural necessity, there were no walls between offices, no walls toward the busy hallways or the offices on the other side of it, and no separation from the open atrium with its noisy canteen and occasional events.
This, we were told, was not an oversight or a cost-cutting measure. It had been a deliberate design by the architect to facilitate more interdisciplinary research. In his mind, a researcher would make a discovery, shout “Eureka”, and researchers from all over the faculty would hurry close, marvel at the new results and start conversations that would spur new, interdisciplinary research directions. It will not surprise anyone familiar with academic work (or open-plan offices) that this was not a successful idea . A year later, the astronomy department paid for walls to be put in around our offices, and everyone’s happiness and productivity went up immensely.
Why am I telling this story, and what does this have to do with data science?
There are data science efforts springing up at universities everywhere. It seems like departments, deans, and provosts have realized that there’s something up with all that data out in the world. There have of course been initiatives that have existed for years now, but I think overall this is a great development. For me personally, because I quite enjoy doing academic research and also like being employed. But also for society more generally, because I think data science research is critically important to many fields, and should not be left to industry alone. There is a unique value in being able to pursue research unencumbered by the economic interests of a company.
It is fascinating to see how different universities and different departments approach the non-trivial task of building data science institutes, for which there is no blueprint. One thing that I’ve overheard multiple times now in different contexts and that I’m somewhat worried about, though, is the idea that any one discipline alone can do it all by themselves. To be clear, I have nothing against focused efforts in a particular discipline. If a university wants to build a focused effort around political science, or physics, or computer science, I think those are all fantastic and valuable initiatives.
But one strength of data science is how universal it is, and how it can be used as a vehicle for building bridges between our extremely siloed scientific domains in order to jointly improve our scientific endeavours across those domains. Paradoxically, there aren’t all that many places in a university where interdisciplinarity is encouraged or facilitated or even just welcome. Interdisciplinary research requires a tearing down of mental walls that is far more difficult than tearing down literal walls between offices.
The data science institutes I’ve worked in are unique spaces in that they encourage communication across language barriers of field-specific jargon and learning of new and unfamiliar things. They value experimentation and unusual approaches. They organize events where researchers can share both results and unanswered questions and brainstorm solutions with scientists from very different backgrounds.
The one common thread tying everyone together is the ubiquitous need to figure out what to do with our data. In my time at the NYU CDS and UW’s eScience Institute so far, I’ve learned about all sorts of fields from the social sciences to oceanography and music technology. I’ve learned from them how they store and share data, how they analyze time series, and what they struggle with. I’ve shared my own experiences in astronomy, and become a much better researcher through those exchanges.
The message of this blog post, then, is this: efforts that focus on one field or topic can be awesome, but don’t loose sight of the fact that there’s an entire world of researchers out there asking similar questions in completely different contexts. Don’t intentionally deprive yourself or your researchers of that rich resource of knowledge, because it could help you make important discoveries that would otherwise be impossible. But whatever you do, please don’t make them work in open-plan offices.