orgtheory.net

the two social sciences

In general, there seem to be to two mindsets in the social sciences. The first I call “precision modeling.” The attitude might be summarized this way:

Social science should focus on simple & clearly defined concepts. Real science is when you formalize these simple concepts into models. The height of empirical research is clear identification of cause and effect mechanisms implied by such models.

The second attitude I call “thick accounts.” Here’s my summary:

Social science should be built around a tool box of flexible concepts. These flexible concepts can be juxtaposed, elaborated and rephrased. The height of empirical research is when researchers can use this tool box to interpret an otherwise opaque complex social domain.

In the first box are folks like modern economists, experimentalists, and generative grammar people. It’s all about seeing first principles and all about tool building. It’s about a style of positivist certainty. A premium is put on tools that extract conceptual clarity. In the second camp, you see qualitative folks, most sociological institutionalists, and historical people. Since formal models usually require simple actors and simple situations, these people can’t stand tool-centric theories that can’t accomadate meaning and eliminate complexity.

Of course, the issue is trade-off. You have to accept generalizability and clarity and dump the overall gestalt. This probably explains the relative strengths and weaknesses of the various social sciences. Economics has good models for how people respond to prices, but they seem to utterly struggle with topics that scream for complexity (e.g., explaining the Great Depression & many macro topics). In contrast, qualitative sciences seem to do better at stuff like discussing cultural change, but I wouldn’t trust ’em with stuff like tax policy analsyis.

I was reminded of this issue while reading James Jasper’s Getting Your Way. It’s a very nice treatment of conflict, negotiation and related topics. On one level, it’s a discussion of the complexity of conflict: you have corporate actors; evolving goals and pay-offs; etc. On a deeper level, it’s a insightful critique of game theory. Game theory’s biggest insights come from understanding relatively simple situations,  but Jasper argues, convincingly, that the actors and games are evolving and complex. You then have a choice: try to build super fancy models or shift to the “thick accounts” approach where you think about the different mechanisms and how they fit together. In sociology, we usually opt for (b), but in economics you move into a world of math, which may have a tenuous relationship with empirical research.

Written by fabiorojas

September 17, 2009 at 12:57 am

21 Responses

Subscribe to comments with RSS.

  1. Kieran

    September 17, 2009 at 1:05 am

  2. That’s a good one – exactly.

    Like

    fabiorojas

    September 17, 2009 at 1:09 am

  3. Methodenstreit

    Like

    Ben

    September 17, 2009 at 2:19 am

  4. The history of science can be seen as a shift towards model 1 from model 2. We knew little about physics until Newton. Similarly for chemistry, biology, etc. We are just starting the trend of using model 1 in the social sciences. Even within economics, you can see the enormous improvements in knowledge within fields–trade theory, institutions, development–that have recently shifted models.

    Another way of telling that you’re in model 1 is if the disagreements between scholars can be settled in principle. Krugman’s screeds in the NYT, and the no-less acrimonious responses, reflect not so much that one side is wrong, so much as the entire state of macro is stuck in a “proto-science” level of development.

    At first, the shift from model 1 to model 2 can involve a destruction in knowledge. Krugman himself discusses how the introduction of modern mapmaking reduced knowledge of Africa, as hazy accounts like “there is a river somewhere there” were discarded in favor of “harder” evidence. But in the long run, rigorous methods were the only way to systematically build knowledge.

    It’s unlikely that we will ever have a (social) theory of everything. The lack of replicable experiments is a big hassle. But the residue of what we can’t explain (ie, philosophy) has been steadily decreasing over time, and is likely to decrease further.

    This would all be much easier if we could study something other than people. Somehow, when it comes to ourselves, we lose all sense of objectivity.

    Like

    Thorfinn

    September 17, 2009 at 3:46 am

  5. kieran, that’s a wonderful piece! from another pov, it all comes down to Hubert Dreyfus and why computers can’t think … or the clash between connnectionist and conceptual models … or btwn statistical learning and rule-following.

    Like

    Tony

    September 17, 2009 at 3:51 am

  6. Yessa. The option to go either route or both routes is one of the reasons I think sociology is superior to economics. Pete Boettke over at The Austrian Economists had a similar post sometime ago urging economists to learn from anthropologists.

    Like

    joshmccabe

    September 17, 2009 at 4:00 am

  7. Thorfinn, I agree that science is a move from Model 2 to Model 1. Fabio didn’t mention social network analysis, but it’s clearly the bit of sociology that is in Model 1 camp, and it’s probably the most successful bit of the discipline right now (there is a reason Social Networks is the #1 subject-specific journal in sociology, by ISI impact ratings; also, I don’t think other bits get 10-page pieces about them in NYT Mag). On the other hand, a premature move from Model 2 to Model 1 is fatal. If Model 1 cannot accommodate all the knowledge and understanding of Model 2 (and at this point it most clearly cannot!) then the extinction of Model 2 research is pure information loss (which happened long ago in Econ, and like is happening right now in Poli Sci). Model 1 needs to chip off, bit by bit, the parts of Model 2 that it can do well. The worst thing that can happen is if Model 1 says that it no longer needs Model 2 (it does), or that Model 1 has nothing to teach Model 2 (it also clearly does). Alas, more often than not, this is exactly what happens.

    Like

    OnNoMoreLemmings

    September 17, 2009 at 4:41 am

  8. @OnNoMoreLemmings

    To say that social network analysis is “the most successful bit of the discipline” is troubling to me (and contradicts the spirit of the rest of your post).

    What does success mean? Is it something that can be measured quantitatively, as the ISI reference suggests? Is it influence beyond in policy? By these measures it is far less successful than neoclassical economics. SO WHAT? That doesn’t necessarily make either of them correct, which surely should be the measure of scholarship.

    As an aside, this strict demarcation between quant and qual seems to be peculiarly American. As does the obsession with network analysis. Why is that? Which methods could tell us?

    Like

    Ben

    September 17, 2009 at 5:18 am

  9. […] des contributeurs du blog Orgtheory.net, Fabio Rojas, propose dans ce billet de distinguer deux types de sciences sociales : le premier correspond aux sciences sociales qui […]

    Like

  10. I’m not sure this distinction should be seen as an opposition. It reminds me what Nelson and Winter called “formal theory” and “appreciative theory” and, as they argued, these are the two faces of the same coin.

    More precisely, if I can think of an appreciative theory without a formal one (but then one can doubt of its progressivity) i’m not sure that there can be a formal theory without an appreciative one. Any formal model has some kind of appreciative theory as its background at least once we ignore the positivist rhetoric of some mainstream economists.

    Like

    C.H.

    September 17, 2009 at 7:42 am

  11. I take science to be the accumulation of knowledge about the world as it really is. If one accepts this definition, which I think most scientists tacitly do in their everyday practice (at least until they get corrupted by fifth-hand readings of Popper or similar), then I don’t think it is at all clear that science is, or should be, inclined towards model 1. Simplicity in concepts and models is only a virtue if the world is simple, but given that such concepts and models in the social sciences are most often achieved by incorporating assumptions that we know are false, I would suggest that there is quite good reason to believe that this is, in fact, not the case. And if the world is not simple, well, then it is not clear to me in what sense model 1 is scientific at all, since it would be in direct opposition to the definition of science provided above.

    Like

    Mike

    September 17, 2009 at 8:59 am

  12. Woops; that should’ve said “production and accumulation.”

    Like

    Mike

    September 17, 2009 at 9:02 am

  13. @Thorfinn
    The last sentence of your post sums up the dilemma quite nicely: Model 1 science is the “better” science as long as you can treat the things you study as objects (like in physics, biology etc.). So being objective (i.e. model 1) in social science means treating people (and organizations, groups, society etc.) as objects.
    This only goes so far, because as we all know, people (that is, “we”) are not objects but subjects, and that is why we also need model 2 research in the social sciences, because it can capture subjectivity.

    Like

    Johann Weichbrodt

    September 17, 2009 at 9:31 am

  14. Analytic Narratives (Bates et al. 1998) immediately springs to mind. Formal models are reductive, yes, but intentionally so. They are a heuristic for understanding complex and contingent interaction and should only be treated as such; confusing the model with the actual events is a mistake (both by readers and authors). To me, it’s the misunderstanding that formal modeling should have an R^2 (so to speak) = 1, when in fact that is not the goal at all.

    The point of reductive theorizing is to highlight the (hopefully) simple causal mechanisms that are operating in a complicated situation and hope to explain a significant portion of variance (figuratively or literally) with a high degree of leverage. Whereas many historical researchers attempt to account for nearly every relevant detail, formal modelers are content with merely the broad strokes.

    Of course, as you point out, different approaches are useful for different social phenomena, but I don’t see how the two are irreconcilable. The disciplinary rhetoric that surrounds both approaches is less useful than identifying the actual reasoning and appropriate circumstances for utilizing one or the other.

    Like

    Trey

    September 17, 2009 at 12:30 pm

  15. Analytic Narratives (Bates et al. 1998) immediately springs to mind

    I remember reading Elster’s review of that book and thinking the authors must have felt much like puppies who have just unexpectedly been kicked by their owner.

    Like

    Kieran

    September 17, 2009 at 1:40 pm

  16. Certainly the project was not without its shortcomings as Elster ably pointed out. I offered it as an example of people attempting to bridge this divide rather than further arging for the balkanization of social science into us-and-them camps in which economists are evil mathematical simpletons and sociologists are the only ones who “get it.”

    Kiser’s review of the book is more sympathetic, though still critical. Strict rational-choice microfoundations based on strategic game theory are not necessary or even appropriate for some explanations. This hardly means that we have to choose one side of the divide or the other. I also don’t find the analogy of puppies and owners fair to anyone involved.

    Like

    Trey

    September 17, 2009 at 5:24 pm

  17. Simplicity and Precision are Two Different Things

    While I agree that the two cultures characterize much of social sciences, Fabio’s characterization conflates simplicity with precision.

    I admire the precision of economics and believe organization theory would be better served if we were more precise in our concepts. At the same time the oversimplification of economics leads us to catastrophe, as witnessed by the recent financial crisis, sustained by simple models for pricing derivatives and simple and precise theoretical models for justifying bad macroeconomic policy (as per Krugman’s recent NYTimes article).

    But abandoning simplicity as an objective need not imply abandoning a quest for conceptual precision. There is too much imprecision in our field, in both quantitative and qualitative work.

    Like

    Willie

    September 18, 2009 at 3:59 pm

  18. […] between two fundamental theoretical approaches that characterizes social science disciplines.  Fabio Rojas (via Crooked Timber) characterizes it thus: In general, there seem to be to two mindsets in the […]

    Like

  19. Sociologists have recently been quite upset that the White House doesn’t want our knowledge or advice. I wonder if the “precision modeling” v. “thick accounts” trade-off has some bearing on that. Are these approaches equally useful for generating policy-relevant knowledge?

    Like

    Cristobal

    September 23, 2009 at 5:43 am

  20. Indeed. People want simple, easy policy prescriptions (e.g., low/high taxes are good), even when some situations (think foreign policy) demand complex prescriptions based on contingencies.

    Like

    fabiorojas

    September 23, 2009 at 8:24 pm

  21. Fabio Rojas, thanks for the post. I found it interesting. But I have to disagree a little bit with some its premises. First, the “tool-centric” approach also uses a series of concepts to simplify complexity and magnify dimensions of interesting. Regardless of whether you are building mathematical models or humanistic language based ones, you are still building models. Each introduces its own distortions. I am always confused when I am told that workers in the “tool-centric” traditions are presenting a “richer” representation of social reality. Is that really true?

    Second, would like to add to this thought . . . “Since formal models usually require simple actors and simple situations, these people can’t stand tool-centric theories that can’t accomadate meaning and eliminate complexity.” What if you are interested in where meaning and complexity come from? The “tool-centric” traditions take them as primitives. Can you investigate their necessary conditions when you assume them? This would be circular. So you would have to approach the problem from other directions.

    Like

    Robert Mahaney

    May 14, 2012 at 12:25 pm


Comments are closed.

%d bloggers like this: