My problematic relationship with theory

I received an email a few days ago asking about a couple assessments I wrote while working for the Army. In those assessments, I laid out an approach to analyzing shared patterns of behavior that I eventually called “Social Terrain Analysis.” When I started working for the government, I saw a lot of analysts going all gaga over things like culture and ideology. I used to study those things too, but eventually decided they’re more trouble than they’re worth: too much folk psychology, too little conceptual coherence. So I developed the assessments mostly to justify my not analyzing those kinds of things, but also to lay out another iteration of a theoretical framework I’d been working on ever since my dissertation fieldwork. 

Some aspects of the assessments I wrote bother me – I was a new employee so I bowed to pressure to fill them with too much military jargon and to oversimplify some of the more nuanced points – but I still stand by the basic assertions I made. The assessments were the last time I made myself summarize my theoretical orientation, and since I feel my thinking has evolved since then, it seemed like a good idea to update that summary. Writing this post, however, has proven surprisingly hard. I think my ideas have changed a little more than I realized. So now this post is an attempt to explain why I have a hard time writing about theory.

I used to be really into theory. My dissertation is basically an argument about social science theory with a few ethnographic vignettes and survey results thrown in for illustration. I sometimes get embarrassed that I made the thing so unnecessarily long, but it was my way of solidifying my early views as a researcher. I’ve provided a link to it because it’s easier for me to point to the references in it than to extract and list all the references here. Anyway, at the time I wrote my dissertation and extending into my early time with the Army, I considered theory the Big Thing social scientists would absolutely need to figure out if the field wanted to get its act together. I don’t think that anymore.

Every theory starts with assumptions, and my main assumption is that psychologized explanations of human behavior have had ample time to prove their empirical worth, and have in almost every case failed to do so (see references in pages 33-42 in my dissertation). By “psychologized,” I mean explanations that refer to some sort of durable internal state that drives behavior. Motives, attitudes, and values are all examples of psychologized explanations. When internal states are generalized to groups, they get called things like identity, ideology, and culture. When they are generalized to groups and about groups, they get called things like tribe, ethnicity, and things like that. Social researchers have tried to use those concepts for…well, for pretty much as long as social researchers have been reasearching, and it’s amazing how inconsistently researchers have been able to explain or predict human behavior when they take those psychologized concepts as must-have components of their research. Most of the concepts social scientists have used don’t have much to show for themselves. They may be useful…maybe…but we no longer need to give them the benefit of the doubt. It’s time to move on.

That was the point of departure I took in my graduate research and in my early Army research, and from there I went on to argue for a specific theoretical framework. The Army version of the framework hinged on the concept of constraints. People of course have motives and attitudes and things like that, I argued, but those thing change all the time in response to all sorts of external stimuli, and similar internal changes can manifest themselves very differently in different people. The consequence of accepting that all of those changes are much too frequent and unpredictable for us to study them successfully, I argued, is that researchers need to approach behavior as if it were just another example of Brownian motion. Go to the Wikipedia article I just linked and look at the graphic of all the bouncing dots. That’s human behavior. A person starts to do one thing for some psychological reason. We don’t need to specify what that reason is because nearly as soon as that person starts doing that thing for that psychological reason, he bumps into reality. Maybe he interacts with another person. Maybe he hits a physical obstacle. Maybe he learns that he doesn’t know how to do the thing that he was originally motivated to do. As soon as that happens, the person changes course, and as soon as he changes course, he bumps into reality again and changes course again. He internal states are constantly changing as he repeatedly bumps into different environmental constraints, which means those internal states are analytically useful for such a short period of time so as to be practically not useful at all. People think and feel all sorts of things. They have all sorts of ideas. They form identities. None of that does us one bit of good from the standpoint of trying to empirically predict and explain durable, shared patterns of behavior.

This view of how people work is consistent with findings from several decades of cognitive science research. I think it would be a stretch to say that that research supports my view – I’m not aware of anyone synthesizing all of those findings into a format that could really be said to support any high-level social theory – but the findings don’t contradict my view, which is more than I can say for many popular social science theories. If you want references for the cognitive findings, here is a classic. If you want specific neuroscience findings, go to Google Scholar and type in “automaticity” or “automatic processing.” In a nutshell, all people’s brains do all kinds of different stuff, but practically all normally-developed, uninjured brains do some things the same: they ingest information about the environment, determine whether that information is tied to something a person would want to pursue or avoid (basically, the degree to which the thing makes a person feel good or bad), and then stores a few sensory cues to remember that aspect of the environment in the future. All of those things operate on the basis of pretty well-established mechanisms, and those mechanism are all I as a social scientist need in order to have a basic explanation of all human behavior everywhere. If human behavior is that moving-dot picture of Brownian motion, the neuroscience explains why the dots move in general. All that’s left to study is what the dots are bumping against, and what the dots do when the bumping happens.

In my Army work, I spent a lot of time working on what kinds of environmental constraints we ought to study. I came up with a list of eight – based on eight categories I developed in my graduate work. I argued that we ought to only look at the ways people can be constrained from being in places or having stuff or knowing stuff or knowing how to do stuff that would make it physically possible to engage in particular behaviors, and constrained from having strong emotional experiences or learning categories or being influenced by peers or developing routines that would make it emotionally comfortable to engage in particular behaviors. My environmental categories were extrapolated from an argument I made in my graduate work about the difference between people and environments and between self-regarding and other-regarding behavior – details and citations starting on page 43 of my dissertation. I think the eight categories are ok. I would say they still needed a lot of work, but I don’t think they do – because I don’t think they matter at this point.

I still think a good social theory needs to address behavior, environment, and thought, and I still think the thought part of the equation should just operate in the background. I no longer think that environmental constraints should be the main focus, serving to predict and explain behavior. I think that should probably become the main focus sometime down the road, but in most cases I don’t think our understanding of behavior is mature enough to do that yet.

I think social researchers have focused too much on explanation and prediction. Both of those things are really important, but they can’t take place until we have a good systematic description of the thing we’re looking at. In my experience, researchers tend to think of description as “qualitative” research – just looking at stuff and trying to make sense of it. But there are a whole lot of ways to rigorously and systematically describe something, and there are times when just having a rigorous description automatically offers up explanations and predictions, and there are times when focusing on explanation and prediction instead of description can actually obstruct understanding.

My preferred illustration of this principle is humours – bodily fluids that, up through most of the eighteen century, were thought to influence people’s mental and physical health. People assumed humours existed based on their assumed effects – excitable people had too much blood, as evidenced by their excitability. The fluids themselves were empirically observable, but their casual qualities were assumed. For centuries, medical scholars ignored and even actively opposed promising lines of inquiry when the researchers who proposed those inquiries challenged the validity of the humour concept. A lot of advances in biology happened after researchers discarded the unfounded assumptions to focus only on concepts they could see. I’m not saying that all not-seeable things should be assumed to not exist – blood carries oxygen to tissues, which is something we could not see until relatively recently in history. I’m saying we should stick to observable things and see how far our explanations and predictions can go just on that basis. We shouldn’t get into theory-building unless we’ve already exhausted most of our potential to just do good empirical description.

Behavioral researchers don’t need tons of theory. They just need to identify some things to observe and then they need to observe what all those things do up until the point that something interesting happens, and then they need to repeat that observation a bunch of times.

I think I avoided this way of thinking for a long time because I assumed – or I wanted to believe – that this was what behavioral researchers already did. I’m not sure of that anymore. One topic Paul and I worked on in our Army jobs was issues of corruption. There is a lot of research on corruption, and relatively little of it seems to focus on observables. Corruption is not a behavior. No one ever pointed to someone else and said, “Look that person’s doing corruption.” Corruption is just a term of convenience that encompasses many different behaviors. When a person pays another person in exchange for a vote, we might call that corruption. When a police official uses his resources and position to intimidate others, we might call that corruption. In each case, the behavior is someone directing some concrete action towards a concrete object or person. All of those things are observable, and making those observations requires very little theory. When we define corruption (or anything else) in theoretical terms and then try to “operationalize” that theory to be able to make actual observations, we create humours. That’s a problem.

There is no way to make observations without assumptions – I’m not deluding myself into thinking it is possible to have theory-less research. And when it comes to rigorous explanation, a robust theory is probably necessary. I think most topics in social and behavioral research aren’t currently understood well enough for rigorous explanation to be a realistic goal. If we take explanation off the table – put it on our list of things we’d like to do eventually but not now – then it’s appropriate to say that the most useful social or behavioral theories are the ones that get out of the way. That’s what I mean when I said I’m a theoretical minimalist.

I think theory can wait. Good explanation may be our ultimate goal, but what we need right now is really good, systematic description. I think the most useful social and behavioral research in the coming years is going to be the research that identifies a bunch of observable behaviors – just a bunch of stuff happening – and then systematically tracks that stuff to analyze what things tend to go with what other things. We can save explanation for later. If we focus on the modest goal of systematic description of observables, we’ll have a better basis for generating good explanation-focused questions in the future. And, if we really do some good, rigorously systematic descriptions, we might be surprised to find that we achieve better prediction and explanation automatically.