Editor’s Note: This is a guest post by Eric Grynaviski, who is an Assistant Professor of Political Science at George Washington University.
There has been a debate on the Duck lately about the meaning of rational choice theory and game theory, and how it’s different from varied alternative approaches (here, here, and here). I wanted to offer a different interpretation than Arena and Jackson. Both give pretty orthodox interpretations, where game theory treats human agents as economic agents interested in maximizing their utility. I wanted to offer a sociologically richer interpretation, concentrating on the idea of how agents react to strategic interdependence.
At the heart of the logic of n-person games is the idea that the best move I make can be affected by what others will do. This is the idea of strategic interdependence. Nexon writes about strategic interdependence that it’s no big deal: “Strategically interdependent preferences of individual actors are not the same thing as a collection of relationally embedded actors.”
The way game theory deals with strategic interdependence gets us into relationally embedded actors. It is a really big deal sociologically speaking.
Interactionism: The way game theory suggests that individuals make sense of interdependent actions is that individuals try to figure out what others are going to do. To figure out what others are going to do, one reasons through what they expect another will do from their position, looking at the world from another’s perspective to figure out what’s going on. Only by putting yourself in another’s shoes can one determine what the best strategy will be. So, for example, in order to figure out how much to offer for a used car, I need to think through what the used car dealer is likely to accept, how often the car dealer wants to go back and forth, etc.
It gets even weirder. The other person, of course, is going through a similar process of putting themselves in your shoes to figure out what you will do in a certain situation. So, you need to put yourself in their shoes as they put themselves in your shoes to figure out what they are figuring out. So, when trying to decide how much to offer, I try to think through how much the car dealer thinks I am willing to pay (this is why I wear crumby clothes to the dealership).
The idea that rational action requires taking others perspectives—that role playing is crucial for social action—is a key concept in much of sociological thought. To take one example from pragmatic thought, think about the overlap between game theory and George H. Mead’s Mind, Self, and Society. The ideas of perspective taking in game theory is the play between the I and the me. For Mead, thinking occurred in a dialogue between my mental representation of what others would do (the me, or others collective attitudes and likely actions) and my response to the me (the I). In other words, what is captured in an extensive form game is the internal conversation of gestures that lies at the heart of Mead’s social psychology.
The same is true in rational choice theory. If I successfully navigate society, then society is ‘in me’ because I can accurately guess what others are going to do in response to my actions; I have internalized the generalized other (and many specific others as well).
Harsanyi’s Assumptions: Less often noticed is a broad set of assumptions made by many game theorists (although not logically required), often rooted in the work of John Harsanyi. First, most modelers (although now there is a debate) adopt the Harsanyi Doctrine that any differences in beliefs must be due to differences in information. Put otherwise, given the same information, we cannot agree to disagree. Second, many modelers assume that players know one others’ cardinal preferences. To make this assumption requires assuming that players with similar social positions understand others valuation schemes. As Harsanyi puts it (PDF), “It is conceivable that in a given society with well-established cultural traditions people tend to enter bargaining situations with more or less consistent expectations about each other’s utility functions. It may happen that all members of a given society are expected to have essentially the same utility function. Or, more realistically, we may assume that at least persons of a given sex, age, social position, education etc, are expected to have similar utility functions of a specified sort” (tug PDF).
This assumption is prevalent in IR work, for example, where there is an assumption that politicians prefer to win office and know that every other politician prefers to win office for example. In fact, whenever a model assumes that agents know others cardinal preferences, it requires making intrapersonal comparisons of utility, invoking some version of Harsanyi’s position.
Now, these assumptions carry really heavy-duty sociological baggage. In order to predict what others will do—to engage in interactionism or to place oneself in anothers’ shoes—we have a shared culture that provides agents the ability to understand others’ utility functions (their interests and how they trade-off against other interests) that is a function of their place in the social structure.
Building on this, we can see how relationally embedded rational choice actors really are. They are only successfully able to navigate the social world to the extent that they have successfully incorporated others relevant definitions of social reality into their understanding of the choice situation.
To reduce rationalism to ‘folk psychology’—the idea that there are beliefs and desires that prompt action—requires a winnowing of the imagination. Rationalism relies on sets of sophisticated sociological assumptions that are compatible with many species of non-rational analysis, and this is before Binsmore and others complicate the logic to make it more ‘social.’
One question is whether these assumptions are reasonable in international politics. I have written about some of these questions; I find the basic setup unlikely to work in international politics because there is too much difference for this perspective-taking to work in a way that produces accurate mental representations of others’ perspectives. But, there is certainly a rich account here that needs to be taken seriously.
Nice piece. I think, though, that you and are using the term “sociological” differently. To my mind, nothing that you say here is actually about sociological embeddedness; instead, it’s about actors inhabiting a rich interactive social context. A relationally embedded actor doesn’t have to “internalize” things, and the whole song-and-dance about taking someone else’s perspective presumes the very autonomous individuality that I am and have been suggesting we need not begin with. Indeed, I read Mead as going through that whole exercise in order to demonstrate the fundamental inadequacy of starting with individuals in order to explain social arrangements; instead, we ought to start with relational/transactions flows and patterns. Which is what I am suggesting we might want to do, instead of building in autonomous individuals at the ground floor of our conceptions the way that decision-theoretic accounts do.
Your strategy seems to be to show how constitutively autonomous individuals are interconnected. I’d prefer to start someplace else.
That’s interesting. I have the entirely opposite reaction to Mead. And, the difference in our interpretation is important. If individuals always rightly interpreted social gestures, then of course the importance of the ‘individual’ drops out strictly speaking because perfect cognition makes cognition uninteresting. Neat features happen when individuals are wrong.
Mead is really interesting in this sense, although he rarely pursues the argument. If gestures are ambiguous–that is, if one can imagine interpreting them as elements of different ‘generalized others’–then one can imagine a case where different agents have different interpretations of their social setting. This happens to me all the time. I go to coffee with a group of people I don’t know, and read the social cues (which are ambiguous) as denoting a friendly conversation, when in fact there is a subtle politics underlying the conversation of which I am not aware (although the others believe I am aware). The key point is that because behavior is ambiguous, allowing a slippage between the social gesture and the interpretation of the social gesture, understanding the individual matters.
This is really quite interesting, because I agree with you about 3/4th of the way, and then you say things like “understanding the individual matters” and I fall off of the bus. The coffeehouse situation you invoke I would describe a bit differently: not that there’s something going on of which you are not aware, which is likely always true, but that the situation as a whole involves different efforts to shape it according to different intentions. Is it “wrong” that you have a different view of things? Is it “right” when your view is brought into accord with the views of those other people? I’m not at all sure about that, and my pragmatist inclinations suggest that the words “right” and “wrong” in this case don’t mean any kind of correspondence between what you think and what those others think, but instead mean something more subtle that can only be sussed out if we know something about the point or purpose of the conversation.
“Neat features happen when individuals are wrong.” I would much rather say: neat features happen when situations are ambiguous, because that’s the opening for creative, contingent agency. Just being “wrong” shuts down agency at the limit called “being right.” Which might be fine for (certain kinds of) knowledge-producing practices, but I am not convinced that it applies to ordinary social transactions.
It’s interesting because I am not quite sure where we disagree. My reading of the coffee house situation is that ‘configuration,’ in the terms you and Dan prefer, of individuals’ definitions of social situations is what creates the ambiguity. The same coffee house situation, without me there, would be different in part because I might be a repository for unexpected and unintended social actions that would disrupt or confuse the conversation.
I realize that this is not identical to the type of analysis of structures that you propose. We disagree I suspect about the importance of fleshing in the social structure with individual-level understandings. The key point though is that this creates a relational theory where the nature of the social structure is produced by individuals’ shifting understandings of the social situations in which they find themselves that conditions interactions in a ‘relational’ way. That is, using the basic set up of rationalist theory, where individuals reason through strategic interactions through perspective-taking (which I take to be nearly axiomatic to rationalist thought), one might develop a relational sociology.
Regarding pragmatic thought, I agree with the nature of ‘right’ and ‘wrong’ here, although I think there is a pragmatic interpretation of this argument. I may develop this in a latter post though, so I duck the issue for now because it requires discussing a different set of issues.
To put it differently, is the ambiguity of the situation described as richly as it might be without describing individuals’ interpretations of the situation?
Nice post. I haven’t read your article yet, but I look forward to doing so when I get the chance.
One quick thought — an awful lot of game theoretic in IR assumes that actors don’t know each other’s preferences. It’s typical to assume they know a lot — almost certainly more than is the case — but fwiw uncertainty over preferences is central to a lot of game theoretic work in IR right now.
Thanks for the comment. I agree about uncertainty. I wanted to state the argument briefly, but the same logic makes sense when there is uncertainty over a set of discrete types. If an agent is uncertain about type, they in fact need to think through how different types would react to a signal,, for example, to identify what that reaction means regarding their type.
Essentially, to translate into the sociological literature, we are unsure of the ‘social position’ (‘type’) of the actor we are engaging with. Through interaction, we use our knowledge of the valuation schemes that agent has (‘the preferences’ for each ‘type’) to discriminate between the types (usually through some sort of signaling game). The backdrop of a social theory to allow actors to navigate society from others perspectives is still there regardless of uncertainty.
Fair enough.