Friday, March 2, 2012

Response to Little by John Levi Martin

[John Levi Martin accepted my invitation to write a response to my discussion of his Explanation of Social Action (link).  John is professor of sociology at the University of Chicago and author of Social Structures in addition to ESA.  Thanks, John!]

Response to Little
by John Levi Martin

The Explanation of Social Action (henceforward, ESA) is a book about the explanation of social action. It argues that there are fundamental instabilities in what seems to be the dominant approach to such explanation in the social sciences, which it terms explanation via third person causality (TPC). TPC is when we attempt to explain the action of persons (say, they vote for a Republican) using entities that are not themselves persons, and attribute causal power to such entities (say, income and authoritarianism cause Republican voting). Examinations of TPC are wrecked if the recipients of the causal force willfully “select” themselves to be exposed to causes. ESA is not the first book to point to the instabilities that result, but previous arguments do not seem to have affected the progression of the social sciences, which operate on the Warner Brothers’ principle that even if you have run off a cliff and there is nothing under your feet, you will not actually fall so long as you do not look down.

As said above, ESA is a book about the explanation of social action—it does not claim to cover everything that social scientists find interesting and important. However, one of the problems with the contemporary approach to the social sciences is that cases of social action are disguised in ways that lead to confusion. Thus Little writes that ESA’s scope fails to include “‘Why?’ questions [that] involve understanding the workings of institutions and structures.” Such questions—e.g., “why did the decision-making process surrounding the launch of Space Shuttle Challenger go so disastrously wrong?”—Little says, “admit of causal explanations. And they are not inherently equivalent to explanations of nested sets of social actions.”

Were this the case, the arguments of ESA would indeed be less damaging. But it is not at all obvious that such “workings of institutions and structures” are not “equivalent to…sets of social actions.” While “workings of institutions and structures” is a fine figure of speech for approximate purposes, speaking more exactly, institutions do not work, or cause, or act, or think. Institutions are nothing other than sets of actions—“patterns of repeated conduct.” When we ignore this, we produce unstable explanations.

Indeed, Little’s example is enlightening. Consider Vaughan’s justly famous The Challenger Launch Decision: Risky Technology, Culture, and Deviance at NASA. This is a wonderful piece of research. But when Vaughan boils things down to a (general) sociological conclusion, it seems of the form “institutional culture caused this decision to be of this sort.” As I am sure a number of people have pointed out, as institutional culture is itself nothing other than the pattern of social action that takes place in the institutional setting, this is to take the pattern of action, pull out one part, put it by itself, draw a line from the rest of the pattern to this one part, and write “causes” on the line. This is a very confusing way of proceeding and formally like the case (treated in ESA) of explaining someone’s favoring of an authoritarian policy by saying that it was “caused” by his authoritarianism. It is taking a tautology and adding a mystical confusion in the form of “causality.” Vaughan’s arguments in detail are important, but we would do better to return to Peirce, James and Dewey and to examine the formation of habits than to posit some new realm of “institutions and structures” with their own causality.

The problems with our desperate belief in causation even where it adds nothing but confusion is best seen in the line of work emphasized by Little on causal “mechanisms.” I did not devote any of ESA to a critique of these ideas in part because I believed that my general points were best made without “going negative,” because I believed that Tilly’s approach to explanation was an improvement over most others, and because of my great admiration for and gratitude to Charles Tilly. This opened space for confusion.

Tilly identifies recurring patterns of social action and calls these “mechanisms” because they can be envisaged as discrete parts of a clockwork that, assembled in a certain way, will produce a certain result. In itself, this is a laudable endeavor; the word “mechanism” to describe the pattern nicely highlights the modular nature of these explanatory nuggets, at the cost of some misleading imagery. But to call these accounts “causal” and to argue that this demonstrates the stability of TPC seems very puzzling. For the mechanisms themselves are nothing other than patterns of action. They do not explain the action, they are it.

For example, Tilly lists “coalition formation between segments of ruling classes and constituted political actors that are currently excluded from power” (henceforward, CFBSORCACPACEFP) as a “mechanism” tending to “cause” democratization. Speaking loosely, this is all well and good, but it cannot be seen as a successful form of TPC. For this requires us to imagine some form of CFBSORCACPACEFP that is forced upon persons as opposed to arising from their own actions (a relation between CFBSORCACPACEFP and democracy interpreted in motivational terms would not support the doctrine of the existence of TPC). Certain cases are more like this extreme scenario and others less like, but no case can perfectly satisfy this vision, because it requires action without action. The more seriously we take the causal language, the sillier or crazier our thoughts become.

In sum, our “run fast and don’t look down” strategy has led us to take outcomes (what happens) and reify them into causes (these are “constraints” and “structures”), and to take patterns of actions and claim that they are not patterns of action at all, but something else that in fact causes these actions. We are akin to meteorologists unwilling to abandon the idea that there really are Winds (e.g., the Zephyr) with their own causal properties, and who abuse any colleagues so simplistic as to insist that to make any scientific progress, they will need to produce accurate models for compressible gasses and then scale up. ESA is about how we might start.


Anonymous said...

First, this reply is a stitch. I very much appreciate the humor.

Second, this reply also seems more or less wholly unpersuasive.

I'll start by saying I've read Dan's initial post, but have not gotten hold of, let alone plowed through, the 400+ pages of ESA.

With that mea culpa on the floor, I'd say that congeries of social actions are themselves complex entities that can produce artifacts (institutions, cultural symbols, etc.) that themselves have independent causal efficacy.

Here is a 'for instance' - Geertz's "webs of significance," the ones agents spin and then get suspended in. That (Weberian) metaphor makes no sense from the ESA perspective as far as I can tell. How is the stickiness and tensile strength of such webs reducible to the individual acts of real live agents?
And, by the way, Geertz was wrong to assert that meanings (embodied in symbols, rituals, traditions, etc.) are not causal. His qualms on that score are plausible only if, like him, we insist that causal account must invoke a general law. Here I think Dan's case (made in his paper on generalization in social science) for the emergence of phenomenal regularities helps us see clear of the error.

Another 'for instance' - the discrepancy between Nash behavior and Structure-Induced equilibria in studies of legislative decision-making. It turns out that "institutional rules" can induce equilibria that are not Nash. Hence they must have some independent efficacy. The rules, of course are artifacts of some earlier set of interactions. So what? They persist over time and do so independently of the initial agents (who might be long deceased or departed from the chamber). If one sticks with rational choice accounts, it is possible to see cheap talk equilibria as representing an analogous example. The emergence and subsequent exploitation of the open-ended rules of natural language has effects that are anomalous form a Nash perspective. In that case too "rules" matter independently of actions.

Example three - In The Public & Its Problems Dewey explicitly invokes the absence of effective institutions as an independent cause of the inability of inchoate publics to coordinate themselves. (Recall his remarks on the persistence of town meeting democracy in a national state and so forth.) And, of course, the reason why modern publics tend to remain inchoate is that they are subject to myriad exogenous factors like demographic change, industrialization, urbanization, and so forth which have enduring, indirect, impact on those not directly party to them. Are such broad patterns reducible to (meaning they are nothing but) the actions of real live agents? My understanding is that ESA invokes Dewey (among others) as an exemplar. My priors are that this would (at best) be a stretch.

There likely are additional examples. Those three came immediately to mind as I read.


Peter T said...

I think the question is sensible. Not so sure I'm convinced by the answer. It's obvious that social patterns do not exist independently of individual humans, so the talk of them as "causes" has to be understood as short-hand for something like "a pattern reliably enough transmitted through various means of socialisation that it replicates across carriers". The trouble is in spelling out the mechanisms of transmission. My tentative answer is that it's not through any single channel, but through multiple, varying ones - language, the natural, shaped and built environments, documents and attitudes to them, education and so on - all working together to pass on to/shape/channel individual humans. the model is language, where no single element carries any element (so "yes" can mean yes/no/maybe depending on tone, linguistic context, social context and so on). There's some interesting work on brain development that's starting to cast a bit of light on this, but not much I'm aware of on how it works in detail in adults.

Perhaps one analogy is a wave in water - each particle is only going up and down - from their POV the lateral motion of the wave does not exist. But the lateral motion causes succeeding particles to move. Do particle motions cause waves?