The Duck of Minerva

Actor Autonomy in Choice-Theoretic Accounts

14 June 2013

Note: this post was co-written with PTJ. Apologies for the comparative lack of structure and the fact that it is a bit repetitive. Note also that it contains a link to a temporarily un-gated copy of Jackson and Nexon (1999). Thanks, SAGE!

In yesterday morning’s post, Phil writes:

One manifestation of this misunderstanding is that “rational choice” or “choice-theoretic” work is often said to favor the agency side of the structure-versus-agency debate.  See, for example, this recent post by Dan Nexon, or the paper it’s based on.  I don’t mean to single my Duck colleagues out, though — the notion that rational choice theorists aren’t particularly interested in structure is quite common.

But I never wrote that “rational choice [sic] theorists aren’t particularly interested in structure.” Rather, we wrote that “choice-theoretic approaches tend to treat actors as autonomous from their environments at the moment of interaction, not so experience-near and social-relational alternatives [emphasis added].” This is a very different claim.

How so? We agree, albeit in a qualified sense, with Phil that:

Most discussion of the observable implications of game-theoretic models focuses primarily on how equilibrium behavior changes in response to changes in structural conditions.  In fact, if one were to insist on committing the error of arguing that a language prevents its speakers from discussing that which they just so happen to rarely discuss, it would probably be more accurate to say that “rational choice theorists” put all their emphasis on structure and trivialize the role of agency.

The following comments focus on applied choice-theoretic frameworks. In principle, such accounts are agnostic about critical inputs; preferences and other dispositions can come from anywhere. In practice, however, this is not the case. For example, expected-utility accounts in political science tend toward locating the source of dispositions in social and cultural structures. One prominent reason is that political scientists usually prefer not to infer preferences from behavior. And it is much easier to specify a set of preferences ex ante if one derives them from, say, an actor’s location in a system of economic exchange or electoral competition.

Indeed, we are pretty explicit about this in the paper. We have a short section entitled “Moving Beyond the ‘Rationalist-Constructivist’ Debate” in which we note that:

Some scholars posit the ‘rationalist-constructivist debate’ as marking a fundamental division in the field. Much of the substantive logic behind this dichotomy involves combining disagreements over thick contextualism and actor embeddedness into a single continuum, in which rationalism denotes a commitment to both thin contextualism and actor autonomy (Price and Reus-Smit, 1998). We agree that such a framing served important disciplinary purposes over a decade ago, but it now suffers from significant problems…. [T]he opposition of rationalism to constructivism follows only from a very narrow reading of rationalism: as a claim that the decision-making procedures that drive human choices are both invariant and also structured by unmediated and objective features of the world.

To circle back to the key distinction: our claim about “actor autonomy” in modal choice-theoretic frameworks is not a claim about the sources of preferences or even decision-making logics, but rather about the model of social action deployed in applied theories. At the moment of action — which is, in this work, the moment of decision — the mechanism at work is lodged inside an autonomous agent. To butcher an old philosophical problem, we might just as well be discussing a brain in a vat.

But this whole line of argument is probably blurring PTJ’s and my attempt to distinguish between philosophical and scientific ontology. In a sense, we’re at risk of arguing about the former rather than the latter. Again, the issue is practical application of choice-theoretic frameworks. Here’s what we say in the paper:

Choice-theoretic approaches include a variety of different theories that explain outcomes by specifying a limited number of actors, their dispositions, and their decision-making logics. Social, cultural, and material contexts supply dispositions and may even explain variation in decision-making logics, but theories themselves treat actors as autonomous decision-makers. Examples include expected-utility theory, many psychological approaches to international politics, and standard implementations of “logics of appropriateness.”

And here’s the table, which should further clarify what we mean and what’s at stake:

Table 1 EJIR

What we describe as choice-theoretic accounts catches up what Charles Tilly called “standard stories” (PDF) while recognizing precisely that structural contexts do a fair amount of work in the underlying theoretical apparatus of these accounts. This is, we think, a feature rather than a bug of our framework: it attempts to sidestep the debate Phil’s engaged with and focus our attention on the ‘paradigms’ associated with explanatory theories… their scientific ontologies.

In a number of rational-choice approaches, the paradigm of social action is what Dewey and Bentley call “self-action” — see Jackson and Nexon, “Relations Before States: Substance, Process, and the Study of World Politics,” European Journal of International Relations, 1999 (temporarily un-gated PDF). This means that the actors are conceptualized as standing independent of one another in a fundamental way, and the source of action comes from within the actors: from their desires, their preferences, their internal dispositions.

The problem, as Talcott Parsons pointed out about 80 years ago, is that one can’t theorize social action understood in this way, so in practice what utilitarians and other decision-theoretical scholars do is to talk a lot about the strategic situations within which these actors find themselves, and reason from situations to outcomes. At most, this shifts applied decision theory into the realm of what Dewey and Bentley call “inter-action,” but remains outside of their third category: “transaction,” in which social action arises not from the doing of autonomous doers, but the configurations of social process out of which actors arise in the first place and keep on arising as their borders and boundaries are (re)inscribed in practice. The point is that decision-theoretic accounts require actors to be constitutively separate decision-makers at the moment of action, regardless of whether their preferences come from purely “domestic” considerations inside of themselves, or from the strategic-interactive environment within which they find themselves.

Consider Erik’s comment on the post in question:

why do you consider actors autonomous from their environment a the moment of decision in choice-theoretic approaches? One popular interpretation presently pursued is that the individual is ‘partly’ free from environmental influences, only in the sense that some part of the explanation for an action needs to refer to the individual and that individuals’ decision. Lars Udehn made this distinction in 2002 in Ann Rev Soc. The latter sense of explanation is quite weak, and does not posit individualism in the sense you propose. After all, the individual is not free to choose their (a) interests, (b) available strategies, (c) who they are interacting with, and in fact do not ‘choose’ their decision because that is provided by the structure of the interaction in combination with the psychological factors, etc. that are relevant to the decision. The interpretation you appear to favor, I suspect, may not be the ‘modal’ interpretation in political science or economic theory more generally.

This line of argument confuses the proximate causal source of individual preferences with the model of social action specified in that framework. And it is in the model of social action that the “actor autonomy” we are concerned with actually lies. Strategically interdependent preferences of individual actors are not the same thing as a collection of relationally embedded actors. Hence, our trichotomous distinction.