institutionalizing computational social science

The following is a guest post by Alex Hanna, a Phd candidate at the University of Wisconsin-Madison.

Fabio’s earlier post on the academic brain drain prompted some good discussion in the comments about students who have computational skills who leave academia for positions in Silicon Valley. Some of the tension in the discussion surrounded whether those students would be better suited for those jobs and how we need those people within the social sciences to handle all the new “big data” that’s coming our way. As someone who’s worked in industry a few times, I don’t exactly think it’s my bag. I’m fairly confident that I’d like to stay within academia. To that end, I want to use this post to think through a few institutional ways that sociology could be changed to be made more amenable to computational social science. By “amenable” I mean trying to incorporate the types of methods and data into the mainstream of sociology research. The exact goals may be a little murky, but a few examples could suffice: publishing big data articles in ASR/AJS or having tenure-track job searches for these types of scholars that are initiated within sociology (and not as a cluster hire or as a search initiated in computer science). I encourage you to add your own below; I’m sure institutional scholars have many, many ideas about this. And I’m sure there’s a lot of fiscal realities that makes all of this sound slightly utopian or maybe even Polyannish. But, taking a cue from Erik Olin Wright, real utopias and so on.

This is also presuming that there’s a critical mass of sociologists that actually want to see the incorporation of computational methods. I know Fabio and Christopher Bail have voiced their support, and there’s that Lazer et al. piece in Science that’s been cited a few hundred times (it’s pretty telling that it was published in a journal like Science), but I don’t know how to gauge this kind of thing outside of my computationally homophilic networks.

1. A Computational Sociology Section
This is a suggestion that socprofessor brought up in the comments of the original article. Some folks wondered if this would actually produce the desired change. I haven’t been around long enough to know about the dynamics of sections and whether they stir up much enthusiasm around their subfield. If anything, it establishes a base of operation of computational social science within the main disciplinary organization.

2. Interdisciplinary collaborations
I know “interdisciplinarity” is the kind of terrible buzzword that administrators love but academics fear. But partnering up with like-minded individuals with different research questions is generally going to be a fruitful strategy. A hard part of this is first getting over the differences in language and jargon. Another hard part is that computer/information scientists may be much more concerned with innovative methods rather than the substantive questions they can answer. And that’s probably mostly where the social scientist’s effort will be needed the most. There’s been some success stories with this kind of arrangement: the Big Data Social Science IGERT at Penn State is trying to hook together social sciences, stats, and CS.

3. Reducing barriers to publication in non-sociology journals/conferences
The norm in many physical sciences, including CS, is to make conferences the main venue of publication, although they do have their analogues in journals like PLoS One (accordingly, a lot of the discussion here will repeat the comments on this OrgTheory post). Conference papers are peer-reviewed, have relatively quick turnaround rates (1-2 months), and low acceptance rates. The kind of research occurring in those conferences speaks directly to myriad standing questions in sociology. For instance, one of the hot topics in political communication at the moment is the focus on political polarization online. But one of the most relevant and well-cited pieces on the topic isn’t published in Political Communication or similar journals, but in a 2011 conference paper by a few folks in the Complex Systems lab at IU.

But there’s a lot of barriers to publishing in these venues. One of them is simply cost. It costs anywhere from $400-600 to register for one of these conferences, which are also often held in lovely international cities. This is, of course, a far cry from the $100 it costs for a student to register for the North America-bound ASA. PLoS One has a publishing fee which is around $1,000. It just often isn’t feasible. There are some institutional entrepreneurs who are willing to take the risk of making these venues more attractive to social scientists; danah boyd has written about the difficulties in doing so for a particular conference. It’s certainly not easy, and it’s something that needs people willing to take risks on both sides.

This is one reason I got very excited about Sociological Science. I envisioned that it’d be a place with a publishing model similar to conference proceedings but speaking directly to sociological concerns. But it also doesn’t solve the problem of being able to engage in other research communities systematically.

4. Considerations in hiring/tenure
Less of a separate point and more of something that ties to items 2 and 3, there doesn’t seem to be much institutional incentives to do the things above. I’m not sure how cross-discipline collaboration is viewed in terms of hiring/tenure decisions, but in any proseminar I’ve sat in on, no one has encouraged collaboration as a means to look more attractive on the job market. And I imagine the same goes for conference proceedings and PLoS One articles.

There’s, of course, a lot I’m glossing over here. Creating new collaborations in a time of reduction of state funds for universities and an NSF under siege by Congress isn’t making anyone’s work any easier. But I hope that there are some feasible moves that we can make to integrate computational sociology as part of the discipline.