I really like computational social science and love to identify myself as a computational social scientist. Having an identity in that sense is perhaps not so important. Throughout my entire career, I’ve got accustomed to being a [one field] person doing [another field] things, and learned to like it. Still computational social science very much sounds like what I want to do myself (at least some of my projects), the only problem is computational social science itself has a bit of identity crisis.
Discrepancies between the literal meaning of a phrase and its usage are, of course, just a common language thing. My bachelor thesis in Chinese was about how the word “kitsch” entered the Chinese language with the translation of Milan Kundera’s The unbearable lightness of being. The translation 媚俗 (mèisú) is (as typical for Chinese) based on the meanings—媚 means “flirt with” or “flatter”, 俗 means “vulgar” and “common” as in the opposite of “refined”. In other words, it’s a Venn diagram, with a bit of overlap between the literal meaning of 媚俗 and “kitsch” but not completely spot on. I argued in my thesis that the use of 媚俗 among intellectuals was influenced by the literal, character-wise meaning.
Just from the phrase, it sounds to me like computational social science is a discipline where one uses computational methods, AI, simulations, etc. to help to build social science theories as a complement to more mainstream approaches. I imagine traditionally trained social scientists sitting down with computational researchers and forming projects from the three questions:
- What are the interesting open social science questions?
- What data can we obtain?
- What methods can we use?
But as Hanna Wallach writes in this great article in Communications of the ACM, the methodology is often instead: “Why not use these large-scale, social datasets in combination with the powerful predictive models developed by computer scientists”. . and see what we get? (I guess you can replace “powerful predictive models” by any (for social scientists) non-standard method.) So “computational social science” has come to mean something slightly different from what it sounds like. I have contributed to this line of research myself, and it is not necessarily bad, but it is hard to reach general conclusions about society in that way. Instead, they will be about the specific medium itself (like that you can infer the strength of earthquakes from twitter data . . which is super cool, but doesn’t help us understand society, or earthquakes (probably they didn’t claim it to be CSS, but it is not inconceivable to see such science at CSS conferences)).
All in all, I don’t think this is a big deal, but in the ideal world, one would separate “computational social science” from “social data science” or “computational media studies”. Not because one discipline is superior to the other, just to resolve the ambiguities of the former term.
To conclude, many people wrote about this better than this blog post. Wallach’s article is one example, Keuschnigg, Lovsjö and Hedström “Analytical sociology and computational social science” from Journal of Computational Social Science this year is another. Also Duncan Watts’s Computational social science: Exciting progress and future directions and maybe even Fredrik Liljeros’s and my “Mechanistic models in computational social science” have more thoughts about how to make successful collaborations between social and computational scientists. I’ll end with a self quote from that paper
To the theoretical natural [and computational] scientists [who want to contribute to social science], we recommend spending a month reading popular social science books. There are too many examples of natural scientists going into social science with the ambition to use the same methods as they are used to—only replacing the natural components by social—and ending up with results that are unverifiable, too general to be interesting, infeasible or already known. While reading, we encourage meditating the following question—why do social scientists ask different questions about society than natural scientists do about nature?
I wanted to wrap this up with some witty reference back to kitsch and Kundera, but given the speed of my wit, I guess I post this now and update the page in a week or so.
2 thoughts on “The mild identity crisis of computational social science”
Thanks for writing this. I especially appreciate the list of references, including the one to your own paper.
The identity crisis you are pointing to results, it would seem, from the attempted merging of very different scientific cultures: that of computational statistics, which is a largely instrumental and engineering-based discipline, and that of the social sciences.
I disagree, politely, with your point that the relevant contrast is between the natural sciences and the social sciences. This contrast is much older that then present one that CSS is wrestling with. Indeed, many debates within the social sciences (such as the viability of positivist research methods) have already internalized the problems with naturalizing social phenomena.
But whereas the social scientists are notoriously fractured into many, many different disciplines with a wide variety of methodological commitments, computational statistics is more cohesive. I have argued that this is due to the advances in the mathematization of the inductive methods used.
What I’ve been looking for for some time is a disciplinary lens that sees computation as a first-order property of societal formation. I.e., how information and data processing complexity bottlenecks contribute to the structure of social forms. Something like that could help resolve the tension between CSS as a bag of tools and CSS as a social scientific field with something substantive to offer social theory.
Thanks for your comments! My background is natural science (with a postdoc and my first faculty position in computer science) thus I don’t want to see CSS as the interface between CS and social science. I agree that computational statistics could help aligning different fields of social science. I’d love to see the outcome of your “computation as a first-order property of societal formation”. I am a bit skeptical tho, as sometimes compute scientists tend to see computation everywhere (photosynthesis is capable of universal computation) . . better just try to tease out the mechanisms. Essentially, I think we can’t do better than taking the research questions / directions from traditional social science.