It is not that they are not modeling “anything” but they are not modeling anything one could actuallly see or observe or test against obervations or evidence.

So, Rawls’s original position and Foucault’s panopticon are “models” – both in their own words and in the terms that, say, Thomas Schelling specifies when talking about models. But neither theorist claims that their model captures the “real world” even as it allows us to talk about things that surely are part of the world – namely justice or power. So too, game theoretic models (that also meet Schelling’s criteria) allow us to talk about something that is part of the world, namely rational action. And that sort of model thereby allows us to talk about the unobservable causal structure that generates phenomena we can observe.

You may want to dismiss that as “armchair philosophy,” but given the paltry showing of the social sciences at generating reliable empirical research I don’t see that as much of a problem. Perhaps the social scientists ought to be a bit more theoretically sophisticated? The problem is that the canonical view of models is that they allow us to deduce and predict with an eye toward testing. That is simply too narrow a view of what models can do or what science involves.

What you seem to be describing is some sort of armchair philosophy – maybe you missed the recent philosophy threads?

To reverse your hammer analogy, there are a great many interesting and useful objects that will not drive a nail. But if you are employed as a carpenter, I think your employer has a legitimate right to insist that you set aside pillows, spools of thread, and butterflies, and employ only objects of proven nail-driving ability in y0ur professional activities.

It seems to me, though, that virtually all of this discussion presumes a rather tendentious conception of what models are for – namely, for deriving predictions or hypotheses that (at least in principle) can be empirically tested. Then we – whether sociologists or economists or whatever – can talk about the truth or falsity of the model.

But if one treats models like tools – things we use to think with – then that view seems at least narrow and very possibly a category mistake. If I think of my re[eated game model as a hammer (or screwdriver or whatever) then there is no reason to think it should resemble the world or be testable against some observations. My hammer does not resemble the world but it is still pretty useful for some sorts of things (and, by implication, not for others).

So, what if we thought about rational choice models as useful for thinking about what rationality means in different sorts of circumstances? Then the models are good for conceptual analysis – for thinking in systematic ways about a class of causal mechanisms and how they work. And that is an important part of the ‘scientific’ enterprise once we give up on a naively empiricist/positivist understanding of science.

Insofar as economists and sociologists (and most of my colleagues in political science) think of model making as an empirical exercise we should say ‘a pox on both (all) your houses).

(In closing I’d like to say hello to my old friend Joe Heath if he is reading this thread too.)

Sorry, if I was confusing. Allow me to try again. First, you are right of course. The paragraph from which the excerpt is making a statement about some methodologies in sociology: it points out that representing society as networks implies that you think society somehow can be fruitfully represented as a network. Put like that, it looks like a tautology but I think this is actually a helpful paragraph, insofar as it clarifies that network analysis should not be construed strictly as an empirical tool.

Now back to morphology. Let me take a concrete example. Say you are a sociologist and you study junior high school classes. You patiently record who plays with whom, who speaks to whom, who sits next to whom in class etc. Then you draw a graph: each vertex is a pupil and each arrow represents a relation. Now, the theoretical insight that permeates network analysis (which the author attributes to Simmel and to Durkheim) is that you will learn something of interest if you study the shape, or form, of the graph you have drawn, or in fancier language, if you study its morphology (the meaning I give to the words “emphasis on the morphological features” of the author is simply “pay attention to the shape”). You could perhaps ask if the graph is connected or else how many connected components it has. Calculating mean distance and/or the incidence matrix, you could propose a formal definition of popularity inside the class. Whatever.

All this is of course really interesting, but some sociologists disagree, or rather have an alternative point of view. Continuing the example, they would say: suppose that a new kid arrives in the class. By your tools and your methods of network analysis, you will eventually be able to determine if he became popular or not, but you won’t be able to predict it, or explain it. This is how you could have done it: instead of observing directly interactions between kids, you could try to write down a list of what kids value. This “list” will in fact look more like a high-dimensional vector space, as orientation is needed because some things that some kids will value highly (say, good grades), some will dislike highly. So you get this vector space of values. Now you try (this I think is the hard part, especially in my example) to plot the kids in this space: Lea has quite good grades, is part of the drama club, doesn’t like Rihanna so much and is very talkative in class so she ends up there in my vector space. So far so good. Then, because you don’t like to represent society as high dimensional vector spaces, you project everything in one or two dimension using component analysis. The, these sociologists say, you will observe that clouds of points appear and they predict that there will be relations in the network you previously drawn between neighboring points. In particular, when that you new kid shows up, you can predict where he will end up, if you happen to know what his inclinations were. The above being a long long explanation of what (I think that) Fourcade meant when she wrote: “When the network approach emphasizes coordination and connectivity among market actors, then, field analysis reveals the topology of social differences.”

In case you read up to there, Tracy, then sorry for having bored you to death.

“And the difference between Anthropology and Sociology is that the former is concerned not only with individual societies but individual persons. Sociology is concerned with people as a mass.”

The anthropologist Alfred Gell argued that anthro was concerned with what happened (and what caused it to happen, and what consequences arose, etc) in the course of a single person’s lifetime, while socio was concerned with the same questions over the course of many generations of lifetimes.

I think the disadvantage with calling these shared norms, images etc ‘social knowledge’ is that it can make you assume that it’s actual facts that are ‘stored’ in these common ideas and institutions, whereas that’s not necessarily the case. For example, the analogy of man is to woman as right hand is to left hand can be used to ‘encode’ an entire ideology of why a gendered division of labour is justified.

Tracy, the paragraph about morphology and formal invariants is really nothing mysterious at all. It’s just a generic way to say that a social institution in which people all connect to each other (which you can represent by a very messy graph) probably behaves very differently from one where all members connect to one central node and that’s all (that one would look like a starfish). The difference in behavior between those two structures would presumably be independent of the culture, history and what-have-you of the members of the network (though as the article points, these parameters would influence how the network came to be structured like this in the first place) and can thus be called “formal”. That this is a question of morphology is simply the definition of morphology. And there you go.