If there is one meta/theoretical debate in Sociology that continually frustrates me it is the dichotomy between agency and structure. Maybe I missed a meeting early on in my sociology education where someone usefully defined them in a coherent way, but these two terms have continually struck me as underdefined, confused, and more harm than help, especially when employed casually and without a rigorous and consistent set of definitions.

A brief example of the kind of thing I’m talking about: As part of a summer program for new graduate students, I attended a presentation on finding good mentorship given by a couple of professors of higher education. The presentation itself was bland but not useless – a few tips here and there about making sure to actively seek mentorship if you don’t happen upon it with your initial advisor, how to make sure you get connected with important faculty so you can find a job, etc. One of the professors leading the presentation then went on to make an extended argument about how you have to exercise agency against the structure of the system. In particular, he meant taking an active role in seeking out resources and assistance from faculty and staff rather than assuming that existing organizational routines will provide adequate amounts without any work on your part.

What strikes me about this example is how un-agentic our grad student agents are. These are agents who are working entirely within the system that is prepared to let them fail, not with the goal of changing it, but rather of getting by. Students are urged to find mentorship where they can, and be entrepreneurial about making contacts with faculty at seminars, in classes, etc. Even though some of the structures in place – like the importance of personal networking with important faculty – seem, well, anti-meritocratic, particularistic, prone to favoritism, etc., students were urged to exercise their agency to take maximal advantage of these structures. I think of this as “agency within structure”, and it’s not the most interesting thing to me – working the system as best you can, but working it the way it wants to be worked. Agency and structure serve here as two sides of the same coin, and not as any kind of dichotomy. Indeed, this “agency within structure” is closely connected to an “agency produces structure” view, where each individual actively reproduces the system in attempting to gain the maximum from it.

More broadly, though, my problem lies with the concept of agent itself. What do sociologists mean when they talk about agency and agents? I do not believe there is anything close to a single answer, but I wonder if we could create a typology of answers. Hopefully someone else has already done this more systematically than I, and some helpful commenter will link me to it. Baring that, I want to suggest a borrowed typology from W. Richard Scott’s text, Organizations: Rational, Natural and Open Systems. In Organizations, Scott attempts to frame existing organizational theory into three broad perspectives that differ fundamentally in how they view the organization.

For example, population ecology approaches might be argued to treat orgs as natural systems*, where each organization is analogized to an organism competing within a niche, trying to survive. The number of organizations in the niche, and external factors, determine how likely each organization is to survive – if a niche is over-filled (say, coffee shops in downtown Ann Arbor), then failure rates will be high. Counterintuitively, pop ecology also argues that in niches where density is quite low orgs will be more likely to fail, as they will lack the legitimacy to access necessary resources. For example, before microbreweries became extremely popular, as they are now, the first might have face difficulty attracting customers and financing, and placing their products in other stores, as there were not systems in place to support that kind of organization – no microbrewery movement, etc. In any case, while not holding rigorously to the framework of “organizations as organisms”, pop ecology treats organizations more like natural systems.

Conversely, transaction cost economic approaches (e.g. Williamson) treat organizations as rational systems designed to achieve a certain outcome**. For Williamson, large corporations formed as became increasingly dominant in modern society (especially in the production of consumer goods) because hierarchical organizations were the most efficient way to organize production of consumer goods. Organizations meet diverse goals – corporations might maximize profits while volunteer after school programs maximize pupils taught – but fundamentally organizations are constructed as rational entities.

The last model is the most subtle, the organization as open system. Here, organizations are recognized as having elements of the natural and rational approaches, but also as being only semi-contained. Organizations are still things – they have a “life of their own” – but they are not closed off entirely, but rather must fight to maintain their identity while maintaining connections with the environment. Contingency theorists (Thompson, Lawrence and Lorsch) are one early example of this approach, as they argued that organizations keep core production processes as isolated from the environment as possible, while devoting other departments almost entirely to interfacing with environmental uncertainty (e.g. contingency). Thus, the production department is not in charge of sales or of purchasing, but rather focuses exclusively on production, with the other two departments guaranteeing a market for the produced goods and access to needed resources for production. More broadly, this perspective begins to break down the barriers between the organization and environment. The phrase “semi-permeable membrane” comes to mind, but maybe I was just paying too much attention in high school Biology.

So, what does this have to do with agents? I wonder if we can break down models of agents even more easily than theories of organizations into the same basic categories: rational, natural and open. Rational models of individual agents are abundant and dominant in many fields, economics especially. Teppo at orgtheory had a recent post that nicely summarized various models of rational agent (although the model “gnat” might more reasonably fit into a natural model here). Agents in these models have preferences and information and beliefs about causality and they go about their business making choices to achieve their preferences. These agents are usually tightly contained (although Teppo’s 4th column attempts to move away from that) and primarily differ in how omniscient they are – perfectly so, in the classic microeconomics arguments, or boundedly so in Herbert Simon’s work***. Rational models argue that we do what we do because it’s what we want to do, and mostly punt on why we might want to do it.

Natural models of human agency are similarly easy to find – from Teppo’s gnat category (which is perhaps a way of putting natural-agent arguments into a rationality formula) to models that assume hunger and libido drive most human action and interaction. Models of inheritance that argue for genetic causes fit in here nicely as well – the search for the elusive gay gene, and the inheritance of intelligence being good examples. The key for natural models is that you do what you do because it’s what you are, on a biological, asocial level.

Open models of agency are a bit trickier, but I think I have one stellar examples. I recently had the opportunity to read The Laws of the Markets****, and in particular, Callon’s introduction. Callon, drawing on Granovetter, argues that we must think of agents not as individuals isolated from each other, acting on each other (as in rational models) but rather as agent-networks. Callon argues that, in particular, calculative agencies are produced through a combination of social and technological forces, and that to understand those calculative agencies we must think of agent-networks using their networks to calculate. Agents do not calculate with the world at large, because it is too big and too remote; instead, agents calculate with their networks. Agent-networks determine what information is relevant or important, how to understand that information*****, what actions are possible, what norms are desirable, etc. Only by conceptualizing the agent as open, rather than closed, can we move away from limited models of rationality.

So, now to return to the initial example. What does this trichotomy do for us in terms of nuancing “agency” in the agency vs. structure debate? Agency within structure seems to me like a rational model – the structures provide the set of possible actions and expected returns, and the agent attempts to achieve existing preferred outcomes by taking advantage of opportunities as best she can. There is no discussion, for example, of how graduate school continually shapes and reshapes both your expectations and your preferences, how your research interests end up looking much more like your advisors, or like what your field believes is novel, or like what funding agencies think is important. An interesting take on that set of subjects is Jeff Schmidt’s Disciplined Minds: A Critical Look at Salaried Professionals and the Soul-Battering System That Shapes Their Lives. I won’t give away too much more than that title does, but let me just mention that there’s a chapter in there about using techniques developed by the army for resisting P.O.W. camps to help resist some of the things graduate school does to you (“agency against structure”, although I am more interested here in his analysis of how the structures tend to shape your agency if you don’t resist them).

Ok, enough on that for today, although I’m sure I’ll come back to it. Anyone have any good suggestions for readings on the broad topic of “agency vs. structure” in Sociology or related fields?

* Ok, I’m not really following Scott’s scheme closely at all, it’s true. Scott uses examples more like Selznick’s to talk about how organizations “have a life of their own” in natural systems approaches.
** Again, Scott uses older examples, like Taylor’s scientific management, to talk about rational systems.
*** I actually think the difference between bounded and perfect rationality is much larger than I make it sound here. The latter is an almost philosophical stance useful in making normative arguments, the former a more realist one that actually attempts a rudimentary model of cognition to make better predictions.
**** This book is absurdly expensive. Anyone have a spare copy they might be willing to part with?
***** Another way to say this would be that agents “frame” their world using their networks. Callon doesn’t draw on Goffman explicitly here, but he does employ the term framing (and a new counterpart, overflowing) as central analytical concepts.