Computational Sociology

[PRivera]

Hi, is anybody pursuing computational sociology? If so, what background do you have? I minored in math undergrad (at an Ivy), and have taken sociology classes, but am otherwise self-taught as far as my computer skills.

For example, I took online classes in Python -- any recs for other languages? Or other things I should be pursuing or learning?

I applied to PhD programs in 2 research areas (computational sociology and an unrelated topic in a related social science). I only got into 2 of these programs, in the unrelated area, but I am MUCH more interested in computational sociology. I'm struggling with whether I should get a Masters at my local, well regarded, big-city public university instead, or if I should just enter one of the funded PhD programs despite not being as interested in the research area. FWIW, one of the two programs I got into is also at a big university with one of the major computational sociologists, but I wouldn't be in his department.

This is a bit of a rambling, "please tell me what to do with my life" kind of post, which I apologize about, but I would love any feedback from anybody pursuing this area... !Gracias!

[electric_burrito]

Hi, is anybody pursuing computational sociology? If so, what background do you have? I minored in math undergrad (at an Ivy), and have taken sociology classes, but am otherwise self-taught as far as my computer skills.

For example, I took online classes in Python -- any recs for other languages? Or other things I should be pursuing or learning?

I applied to PhD programs in 2 research areas (computational sociology and an unrelated topic in a related social science). I only got into 2 of these programs, in the unrelated area, but I am MUCH more interested in computational sociology. I'm struggling with whether I should get a Masters at my local, well regarded, big-city public university instead, or if I should just enter one of the funded PhD programs despite not being as interested in the research area. FWIW, one of the two programs I got into is also at a big university with one of the major computational sociologists, but I wouldn't be in his department.

This is a bit of a rambling, "please tell me what to do with my life" kind of post, which I apologize about, but I would love any feedback from anybody pursuing this area... !Gracias!

I do computational sociology/social science so here is my $0.02.

 

Learn R. It is free, well-supported, extensively-documented, and pretty much the industry standard in computational social science research. Just browse through r-bloggers.com to see how widely-used it is. Sure, it has its problems, eats up your memory and takes its sweet time analyzing more complex models, but with cloud programming becoming the standard soon that won't be an issue anymore.

 

On top of that, I'd say learn Python, Stata, and SAS too. In my experience, different fields have different preferences. Social science people typically use R; public health like SAS. It just depends on what kind of research you want to do.

 

Try to learn as many statistical packages as you can. You'll find that knowing different languages/programs will help you avoid certain problems and glitches associated with other programs. R is notoriously bad at recoding variables, so I recode in Stata and analyze in R. Also, regardless of your field, having programming skills AND a phd degree will almost guarantee you a data analysis job almost anywhere. So, it's a great way to build a secondary skill-set in case academia isn't your thing.

 

I'm happy to discuss further anything related to code/computational methods.

Edited March 26, 2014 by electric_burrito

[RefurbedScientist]

I'll try to give a little feedback to your bigger question about phd programs. 

 

As you probably already know, there isn't a lot of infrastructure in place to support computational sociology, yet. That doesn't mean you shouldn't go into it, but you'll be hard pressed to find a program that bills itself as "strong in computational sociology". Rather, you'll find individual faculty here and there who have those skills. Even more likely, you'll find clusters of grad students working interdepartmentally with comp. sci. or other fields. 

 

The point is, you would be ill advised to market yourself as a budding computational sociologist and nothing else. You'd want to bill yourself as a sociologist of stratification/race/media/politics/urban/whatever with an interest in computational methods. If you want to go into a traditional sociology department, then that's the language you're going to have to speak.

 

Now, if you don't want to go into a traditional sociology department, then that's another issue altogether. Maybe you want to be the next Duncan Watts. Maybe you want to do big data analysis for some urban planning. Maybe you want to make sweet info-graphics. For all of these things and more, you probably don't need to get a PhD in sociology. If you can pick up a MA for cheap at a public university and get a job in industry or government that pays well, then skip the PhD. If you really want to be a professor but don't have any substantive sociological interests, then maybe that other social science field is for you. 

 

And if you're not sure about all of this, it never hurts to wait a year. In the meantime, pick up R and sharpen your Python. Certain departments (and not others) will really appreciate these skills. If you're committed to sociology, then I would say skip the masters and forgo the other PhD program. Spend the year sharpening your statement of purpose (with Grad Cafe's help) and targeting your applications better. If I were to guess (knowing nothing else about you ) , I would assume your GRE's were fine but maybe, being a math major, your SoP wasn't "sociological" enough. The problem with being on the cutting edge is that you're always doing extra work to clear the path for others to follow, so keep it up!

[jacib]

I am not a computational sociologist and have no career advice, but I can tell you my friends who are interested in computational sociology all read the blog Bad Hessian. 

[gilbertrollins]

The point is, you would be ill advised to market yourself as a budding computational sociologist and nothing else. You'd want to bill yourself as a sociologist of stratification/race/media/politics/urban/whatever with an interest in computational methods. If you want to go into a traditional sociology department, then that's the language you're going to have to speak.

 

This is great advice.  If you like writing programs and proofs because of the elegance and gee-whiz factor, then you're going to need to get connected to some of the debates in sociology, economics, or political science, if you're going to import computational methods, because it's a struggling subfield in all three of those disciplines.  Economists are not impressed with agent based modeling because there have not been any really huge breakthroughs (in terms of solving fundamental problems in the mainstream), and agent based models require way more assumptions about agents than does traditional modeling in econ, where parsimony still rules.  I can't say a lot about political science, but they're still catching up to econ in terms of formal theory.  There is better reception for quantitative work, i.e. more of it and more people who understand more sophisticated levels of it, in general in political science than in soc, though.  

 

In soc there is a small and vigorous crowd of math and physics orphans who got bored with pure logic and wanted to switch into something that is more applied to people and all their big feels.  There is a lot of solidarity among these people, but their main concern right now is how to get the stuff into some relevance in the mainstream of soc.  Outside some statistics, sociologists are not fans of thinking of behaviors themselves in formal terms.  Rational choice died a slow and painful death in sociology, and people still stop by its grave to take a piss every now and then.  If you're coming from the quant side into soc, your biased sample will make it look like sociologists do a lot of social networks stuff - and that is incorrect.  There is, again, a nice solid crew doing networks in sociology, but it is a somewhat nascent subfield.  Most people who talk about "social networks" in sociology are just using it as a techno-jargony way to refer to more traditional sociological theory.

 

One option for you is to get yourself in a computer science PhD program.  Those guys can pretty much go wherever they want and do whatever they want, is the impression I get, and they have tons of options in the non-academic job market.  Doing that and getting some crossover in social science would put you in a really good position for private employment doing data science.